I’m closing out the school year by slowing down to actually look back and make sense of what happened. It’s a metacognitive act that I, for one, could certainly do more of.

In this article, I’m inviting a few guests from the podcast episode “How Did It Actually Go?” (link below) to guest-write an article. Each of them agreed to go a little deeper in writing, reflecting on how process-based learning with AI actually played out in their schools.

In this article, Leon Lam reflects on building an AI chatbot for process-based learning and what he’d do differently next time. Reflecting back, what he found is at the center of this piece, so I won’t spoil it. But here’s a takeaway I want to reinforce: he assumed students would follow the process without ever explaining why. A process that remains invisible asks students to comply rather than learn. Adolescents need both to understand why we do something and how it helps them grow.

Let’s check in with Leon to hear his thoughts.

Intro

Leon Lam here, guest posting on Alex’s Substack. This will be an extension of what I discussed on the podcast episode “How Did It Actually Go?” I want to dive deeper into what I learned building an AI chatbot for process-based learning, and what I would do differently next time.

A lot of teachers use custom chatbots in their classrooms. I went a step further and built an entire platform for creating custom chatbots, mostly for analytics and data. I wanted to know whether or not the AI chatbots were making an impact. My hope is that what I learned can help you decide whether to invite an AI tutor into your classroom or to keep it outside.

What I Built

As mentioned in the podcast episode, my Socratic essay-writing bot coached students through Cambridge AS Economics 12-mark essays. It was structured around stages: question analysis, planning, and paragraph coaching.

In my bot, Alex’s Think, Generate, Edit process was built into each stage. Students had to think through the questions the AI gave, craft their own responses, which the AI gave feedback on according to preset criteria, and then they had to edit their work until it matched the criteria. The stages were important because they allowed students to plan first instead of jumping into the writing immediately.

What I Observed

I categorized student behavior into three patterns. The first group of students spent hours with the bot. They exchanged hundreds of messages. They followed the bot’s strictly enforced reply format. They complied with it for as long as it took to complete the essay on the platform. It looked like they were fully engaged, but upon deeper digging, I discovered that the chatbot I made was needlessly ruthless, and that the learning could’ve happened much more quickly if I had built in human touchpoints or relaxed the restrictions. Students had reported how cumbersome it was to get the exact answer the AI wanted.

The second group of students tried to game it for answers. The bot was designed to ask them questions, so they worked at getting around that. These students weren’t really learning anything. They tried prompt injection, off-topic detours, anything to extract an answer. They mostly failed, but time was wasted on trying to manipulate an AI instead of learning.

The third group of students did not engage with it at all. That was just not how they wanted to learn. In the end, the bot was just a chat interface. They wanted a teacher, and I saw their eyes light up when I took back the reins in the classroom.

Performance on summative assessments did not change since using the bot, either. Students who did well before continued to do well. Students who struggled before kept struggling. What did change was my ability to see the process. I could zoom in on student artifacts and query the data with an LLM, and that visibility was genuinely useful. So in that way, you might say the bot served as an assessment, which provided data that in many ways reinforced what I already knew about the students.

I suspect the issues came from two main reasons:

1. I did not name the thinking processes out loud with the students. I built it into the bot and assumed they would just follow along. Thinking back, I should have made them aware of the process so they know why I chose to set up the assignment this way. My thinking is that it would help the hacker group and the disengaged group to want to use the bot meaningfully.

2. I removed too much of myself from the process. In theory, the assignment should have worked. In practice, my students handed the entire feedback process over to an AI, and the efficiency I was chasing became the flaw. Aimée had named it before I did. She’d been a guest on the same podcast; when I heard her portion of the episode, her idea of “gates” mapped exactly onto what I’d been considering ever since I put the bot in front of students.

What I would do differently

I do believe that process-based learning is the way to go in the age of AI, but my next iteration of this assignment will definitely be different. Here’s how I will pivot moving forward:

I will reinsert myself in the feedback loop. AI should not give feedback on its own, no matter how well it is trained. Giving feedback to students is what builds trust and rapport; that’s the teacher’s job. There’s still a real role for AI, though. An AI can be trained to spot specific writing weaknesses and tag them to feedback I’ve already written, or to extend a comment of mine by pointing to a resource. The condition is that everything passes under my eyes before it reaches the student. The main point is that AI-use should reinforce things that we are learning and want to support in class.

I’m going to teach the process out loud. I will explicitly teach my students Think, Generate, Edit, and other processes, or co-create a process with them in class that suits the task. This time, they got a tool that already knew the answer to that question, and they were left to comply with it. However, if I had printed our process on paper and written underneath each step where and how AI was used, and why, my students would have understood the design from the inside and hopefully have been engaged with every step.

I’m changing how I grade. Because some students will focus too much on the final output, even if I ask for process artifacts, I won’t accept a finished essay unless the artifacts back it up. The artifacts will be graded, too, but the bulk of the grade for the final product will be awarded only if the final output was the natural product of the process.

I’m magnifying authentic, in-person assessment without neglecting AI literacy. I want students to share opinions that are actually theirs, in front of other humans, with their screens closed. That is often uncomfortable for them, and that is the point. My ideal classroom is one that maximizes original thought, critical thinking, and other capacities needed to interact effectively with AI, which I am making a focus outside of the classroom. Students will be instructed to interact with AI without my supervision. This means I will need to teach AI literacy, so my students can remain thoughtful and responsible.

I’m still building, but smarter. I built a platform that made students learn through Socratic chatbots, a workflow that’s still new and unproven in their minds, piled on top of everything else they already had to do. So the next iteration starts from what students already do instead of inventing something foreign. I rewrote the entire textbook with AI for accessibility, same flow through the topics, but with simpler, more direct wording. I put MCQs inline for formative checks, and digitized over 3,500 past-paper MCQs organized by topic so students can set up their own mock tests, complete with explanations for the wrong answers. Because I own the platform, I’ll know exactly which topics a class struggled with. So I will keep building my own platform and iterate on the processes students already go through, improving their learning and my teaching at the same time.

Conclusion

Thinking back to my own time in schooling, I don’t remember which teachers had the coolest PowerPoints or used the latest gadgets. I remember the chance to question alongside other students, the jokes, the moments of wonder and care. I miss watching students struggle through the thinking, take pride in what they made, and own their learning. With the changes I’ve shared, that’s the classroom I want to build in this AI-enabled world.

Monday Ready Resources

One of the most important realizations from this experience is that I need to talk to students along the way about the process or even co-design it with them, as well as teach AI literacy. Here are three ways you can get started with your students to make a process or build AI literacy:

1. Go to Alex’s AI Enhanced Process generator and take a crack at making your own process on your own or with your class.

2. Take Anthropic’s free course on AI Fluency: Framework & Foundations. This can also be done on your own or with your students.

3. Use my Process-Based AI Use Scale which you can post around the room, or use with students to determine how much AI should be used in each specific step of the process.

AI Disclosure

I wrote this whole article, and then Claude was consulted on for wording, structure and flow. Some of Claude’s suggestions made it to the final version. Some suggestions inspired other original changes. Ultimately, the words are entirely my own and represent my opinion. Aside from the screenshots of my application, I gave ChatGPT the final version of this article, asked for image suggestions, and asked it to craft the prompt that generated the images you see in the article.

Images generated by ChatGPT and Gemini.

Intro

I’m closing out the school year by slowing down to actually look back and make sense of what happened. It’s a metacognitive act that I, for one, could certainly do more of.

Over the next few weeks, I’m inviting a few guests from the podcast episode “How Did It Actually Go?” to guest-write an article. Each of them agreed to go a little deeper in writing, reflecting on how process-based learning with AI actually played out in their schools.

Writing is both output and thinking. Writing is the actual process of figuring out what you believe. I’ve found this idea to be true on this Substack, and I think my guests have too. There’s a depth in that kind of deliberate slowing down that I haven’t always experienced with AI-generated text. Personally, I can’t help but wonder whether that reflective habit is at risk. The world is moving fast, and every week there’s a new exciting model. Deliberate slowing could look inefficient in that context and anti-zeitgeisty.

With that, I want to thank Aimée Skidmore for this week’s post, which sits at the center of this discussion. Aimée teaches Grade 12 in Geneva and thinks hard about when GenAI is in the room, how we can maintain student agency and effortful thinking, as they are prone to wanting to move too fast in the name of completion. To get students to deliberately slow down, she created something that she calls gates that serve as pauses, or deliberate moments in a thought process, where students have to show what they are actually thinking before they earn the right to move on. The idea came from the game Dungeons and Dragons, which tells you something about how Aimée thinks. She’s practical, a little playful, and genuinely curious about the tension between structure and ownership in a classroom where AI can skip the messy middle entirely.

In this piece, she walks through two iterations of the same project, what she noticed between them, and the questions she’s still thinking about. The Monday-Ready resource at the end is concrete and immediately usable, with a checklist of things we can do with gates in a process.

Make sure to follow Aimée on Substack. Enjoy!

From Aimée

When students use GenAI, the worry is that they’ll outsource the final product. But the bigger risk is that they outsource the messy middle: the testing, rejecting, revising, deciding, and explaining.

So many of us avoid this issue by designing around GenAI. And I used to spend a lot of time wrangling with how to do this with some of my lessons and projects. Now, I spend less time doing that and more time engineering moments where students have to show what they are thinking before they move on.

My Grade 12 students were working in pairs to build a chatbot to help another student practice a certain habit of mind, like persistence or thinking flexibly. I wanted them to work through a Design Thinking process of empathy, define, ideate, prototype, test. Some of these steps involved getting support from GenAI, and some were not. I wanted them to be balanced in their use of tech.

Alex McMillan’s AI Enhanced Process Generator was a key tool in helping me decide and communicate on which steps students might use AI to help and where I wanted them to work on their own. Full product scrolling screenshot below.

At first glance, it could have looked like a dream GenAI project. Students were using AI, building something for a real purpose. They seemed to be in the flow and moving quickly. Maybe a little too quickly.

I started noticing that students were at their computers, starting to build the chatbots, pretty early on. Some were even submitting the link to their final product in one class period. I felt a little panic and then decided to walk around and ask how things were going. What I found was disappointing: I couldn’t get to every student, there were some who couldn’t answer my questions about their process, and there were some who didn’t accept my suggestions to slow down and have another look at the first steps.

So I went back to the drawing board to rethink the approach and rebuild it for the next cohort. How could I get them to slow down and go through all the steps of design thinking? I was trying to find out how I could get them to hand in a ‘rough draft’, like we do with essay writing, but I was more interested in checking their process than their product. I didn’t really care so much about whether the chatbot was 100% functional. It was only one small piece of the project rubric.

Iterating with Gates

On my second iteration of this project, I decided to add some proficiency checkpoints: a pause and check that students have to take before they move to the next stage of the work. I called them gates because I had this image of a DnD player facing an important decision where they need to slow down, check equipment and consult with their party before going through.

Here are the two I built:

Here’s what happened:

The pace slowed. Students appeared to be more thoughtful in their choices. They had to sit through the struggle and check their own work before asking me.

The talk changed. I was able to have short conversations with each student when they called me over to sign off. Over time, our talk became less about me checking their work and more about “Tell me where you are now.” “What do you like about this tool so far?”

Students started explaining choices. “What led you to that decision?” I was able to redirect them when I saw they were not thinking deeply enough and ask them some questions that made my coach’s heart flutter. “What was challenging here for you? And what else?”

They noticed problems earlier. Before they handed it in, they were able to make improvements because they could see those changes would make the final product stronger. The project became less about “my chatbot works” and more about “my chatbot is designed for a real learner.”

This felt like a real win. The gates did what I hoped they would do. They slowed the project down in the right places. They made the process more visible and gave students a reason to explain their choices before rushing ahead.

And this is the part I’m still thinking about. I feel a tension here about how much of the process I should define for them.

When I create something, I do not move through the work in a straight line. I start in one place, jump to building, get stuck, jump somewhere else, come back, revise, test, rethink, and slowly find my way through. That movement feels natural to me now, but it took years to build. Students are still learning what that kind of process feels like.

So the questions I’m sitting with now are: how do we give students enough structure to support their thinking, without turning the process into another set of steps they simply complete for us? How do I avoid a heavy process that will lead to more paperwork and overfunctioning for me?

Because if I build too many gates, or if every gate depends on my approval, I risk creating the very thing I’m trying to move away from: students waiting for me to tell them if they are doing it right, if they are allowed to continue.

So, the next version of this project might have students deciding where the gates go. It might involve more student self-checks, more peer testing, and more room for students to say, “This is what we tried. This is what we changed. This is why we’re moving forward.”

And probably more modeling from me, too. Not modeling the perfect process, but showing what it looks like to get stuck, change direction, reject an idea, return to an earlier version, and keep working. That feels important because students do not learn ownership by being dropped into total freedom. They learn it by practicing responsibility within a structure that helps them keep going.

The gate is not the point. The pause is the point. And what students do inside that pause is where the learning lives. That, to me, is one of the real design challenges with GenAI in the classroom. Yes, the tool can make the work move quickly. My job is to help students slow down enough to notice what they are doing, make real choices, and stay awake inside the process.

Monday-Ready Resources

Resource #1 - Checklist when Using Gates

Separate the gate from the grade.

If students associate checkpoints with judgment, they’ll perform readiness rather than demonstrate it. Frame the gate as a conversation. “Walk me through your thinking” lands differently than “let me check your work.”

Unpack the steps before students take them.

When you introduce a process, explain why each stage exists. Human psychology is consistent on this: we do not expend effort on things that feel arbitrary. If students understand why the empathy phase comes before the prototype phase, they’re more likely to take it seriously.

Use a student-facing checklist, then release some gates over time.

Before students call you over, they should be able to say yes to two or three concrete criteria. This shifts the first layer of accountability to them and changes what the teacher conversation is actually for. Over time, some gates can become peer-checked or self-certified. Early on, every checkpoint might involve the teacher. Once students show they understand the process, they can take on more of the checking themselves. This builds toward ownership without dropping them into total freedom before they’re ready. You can see how I built this into Step 5. Test on the Project Worksheet. (link below)

Create a process journal and build in feedback before moving on.

Ask students to document their thinking at each stage before they call you over. The journal becomes evidence of the work, not just the product. A peer can respond first; the teacher becomes the second reader. You will see how I did this through a Project Worksheet. (link below)

Practice the process more than once.

Research on habit formation and classroom routines suggests it takes roughly three iterations before a process becomes something students internalize and implement with any real fidelity. The first run is orientation. The second is where it starts to click. The third is where it becomes routine.

Go public with stuck moments.

Model a process where you make a wrong choice, back up, and explain why you changed direction. Do this more than once. If the only process students ever see is the polished version, the messy middle feels like a mistake instead of a sign the work is actually happening.

Resource #2 - Printable Checklist

Here’s a PDF you can print out, along with look-fors, for using a gate in your processes with students.

Gates Checklist

167KB ∙ PDF file

Download

Download

AI Disclosure

ChatGPT was used to help me code some pieces for the Teacher and Student Project Brief. It also helped check the project for safeguarding issues, and gave the idea to have students think about and build in guardrails. I used it to generate the Project Rubric.

Image of the gate created with ChatGPT to represent the idea of a gate: a pause where students make their thinking visible before moving on.

Elements of the article leveraged AI as a supporter of language clarity, not idea generation. All conceptual content of this article was created either by Aimée or Alex, or both in partnership.

The Project Links

Check out the project here referenced in the article.

• TEACHER version - https://buildhomassistantteacher.netlify.app/

• STUDENT version - https://buildhomassistantstudent.netlify.app/

It's time to finish up the year with one last podcast episode. I decided that I wanted to have a reflection and talk to people about how process-based learning has been going inside their schools or classrooms. I talked to a range of educators and asked them several different questions, and this episode is a series of highlights from those conversations. So, over these 20 minutes, you're going to hear a series of short recordings in which we look at process-based learning with AI from several angles. Below are notes about each of the guests with links to their websites and social media. Thank you all for contributing to this episode!

Aimée Skidmore | Teaching and Learning Coach | Geneva

Aimée works with experienced teachers who are tired of being the engine in the room. Her focus is student ownership: structures where students start, think, revise, and take responsibility without the teacher carrying it all. She appears twice in this episode. First, she describes what process-based AI use looks like from inside her classroom. In her second segment, she explains how deliberate checkpoint gates changed the outcome of a chatbot-building project.

Aimée offers a six-week Student Ownership Sprint for secondary teachers. She also hosts the International Teacher Staffroom podcast.

LinkedIn | TeachSpark

Aimée wrote a companion piece to go along with this episode. After you listen, make sure to read her more in-depth write up about “gates” below.

Jay Goodman, Ed.D. | PBL Consultant | Canada

Jay has spent nearly two decades designing problem-based learning programs. His Ed.D. focused on PBL program design. He co-developed the Innovation Institute, an award-winning interdisciplinary PBL program in Shanghai.

In this episode, he describes mentor bots: teacher-designed AI personas built around specific domains of expertise. Students identify a knowledge gap, do initial research, and then bring that thinking into a structured conversation with a field-specific model. It solves a real PBL logistics problem without replacing the thinking students need to do first.

LinkedIn | Goodman Learning Partners

Vamshi Mugatha | Director of Technology | American School of Brasilia

Vamshi brings in a leadership perspective as an admin. Vamshi describes a familiar challenge for many schools around the implementation side of a policy. What he realized was that the missing piece was expectations. When teachers weren’t setting them, students were using AI without disclosing it. The gap between the two created tension that the policy alone couldn’t resolve.

LinkedIn

Leon Lam | A-Level Head of Humanities | Beijing National Day School

Leon teaches A-Level economics and leads Humanities at Beijing National Day School. Last year, he vibe-coded a Socratic essay coaching chatbot designed to slow students down and move them through idea generation, outlining, and drafting as distinct stages. He’s candid about what happened. Some students engaged deeply. Others focused entirely on getting the chatbot to advance to the next stage, treating compliance as the goal. He reflects on what he’d do differently next time. His biggest takeaway is that co-designing a process with students can be a powerful way to make the process less performative and more purposeful in supporting their work.

LinkedIn

Check out these 3 posts with photos from previous AIFE conferences.

This is one of a growing collection of AI-enhanced processes I’ve designed to support different types of thinking in the classroom. You can find the full collection at aienhancedprocesses.com/t/processes. I’ll keep adding to it as I build more. The content below is a preview; the full collection of processes will be available in my book.

At a Glance

Hindsight, Oversight, Foresight treats metacognition as a practice distributed across the entire learning experience. Students can start practicing metacognition at any point, revisit steps, and cycle through them as the work demands.

• Hindsight looks back to understand what happened and why.

• Oversight pauses mid-process to notice when thinking is stuck, drifting, or needs a new direction.

• Foresight scans what’s ahead before diving in.

AI can support any stage or none; where it fits depends on where a thought partner would help without taking the thinking away from the learners.

Hindsight: Reflecting on Events

After learning, we look back to understand what happened. Hindsight helps us notice what worked, what didn’t, and what to change next time. It’s also when we connect outcomes to the choices or strategies that caused them, so learning can transfer.

Below are the three steps in this process.

1. Highlight. Students identify what stood out, including what worked, what was difficult, where things went off track, and surprises they encountered. AI could be used by asking kids about the event and drawing out their thoughts.

2. Hypothesize. Students think about why events happened. They consider causes, choices, misunderstandings, strategies, or conditions that contributed to the outcomes of the experience. AI could support by asking questions to see cause and effect relationships.

3. Act. Students capture what they want to take away from this experience. It could be an insight, a strategy note, a checklist, or a reminder on how to work with a team member, etc. AI could help summarize, label takeaways, or suggest transferability. It could also prompt learners to consider their new learning and how it could be applied in the future.

Oversight: In The Moment Awareness

Oversight is that internal version of formative assessment. It is noticing when we are stuck, distracted, guessing, or meeting our goals, and identifying what we might need next. It creates the pause to ask whether our current approach is working and whether a shift in strategy, pace, or resources could help.

Below are the three steps in this process:

1. Observe. Students pause during a task to scan their thinking and emotional state. They identify signs of confusion, distraction, guessing, confidence, or progress. AI could be used to ask students to check in, notice, and draw their attention to what is happening internally and externally.

2. Organize. Students make sense of what they noticed by sorting, interpreting, or connecting their observations. AI can help them clarify patterns, name what might be happening, or rephrase their thinking without solving the problem.

3. Act. Students choose the next step and what to adjust. AI could help them weigh options, break down steps, and so on; however, decisions ultimately stay with the students.

Foresight: Planning for Success

Before learning begins, students who pause to anticipate challenges, activate prior knowledge, and identify useful strategies are more likely to start with intention. They’re also less likely to stress out when things get hard. That pause is metacognition, and it turns out, doing it before you need is a powerful practice.

Below are the three steps in this process.

1. Find. Students scan the task, prompt, or situation and identify key elements like requirements, keywords, constraints, or prior connections. AI could be used to ask students about the task, helping them to break down assignments into the above parts.

2. Focus. Students surface questions, possibilities, or uncertainties about the elements in Find. AI could be used to support their curiosity and reveal what thinking or clarifications may be needed before action.

3. Act. Students choose a direction, where to start, what strategies to try, or what support resources to use. AI could be used in this step to support thinking through questions and paraphrasing; again, students should make all decisions and AI supports their thinking to do so.

If you would like to share how you used AI and metacognition, leave a comment below. I’d love to hear your story!

Thank you for reading!

Scaffolding is one of those practices most educators have been trained to use, talk about as a part of daily planning, but might need to reconsider now that we live in the age of AI. We’ve been using it for a long time: breaking down a complex task, modeling a thinking move, offering a hint when a student gets stuck, then stepping back as they find their footing. But knowing what scaffolding is and implementing it with fidelity in an AI-enhanced classroom are two different things. When the support is too tight, too scripted, or never fades, scaffolding can stop being supportive of student learning and growth. In a classroom where AI is available, a student oriented toward completion rather than understanding is one click away from outsourcing the whole thing.

So this article is about one question with two parts: what does strong scaffolding look like when AI is in the room, and how do we design for it deliberately? The research on effective scaffolding gives us the foundation. AI gives us both a powerful new tool and a new set of risks. Understanding both is what makes the difference between AI enhancing student thinking and replacing it.

Definitions

I find that Universal Design for Learning (UDL) and differentiated instruction (DI) are often used interchangeably with scaffolding, so I want to take a minute to explore the three of them in relationship to one another before we move forward. I imagine that the conflation comes from the fact that each involves a teacher adjusting something for a learner to be successful. But there are some nuances between the three, and to make it more interesting, all three could technically exist at the same time.

Defining Universal Design for Learning (UDL)

CAST, the organization that developed Universal Design for Learning, describes UDL as a framework for designing curriculum so it works for all learners from the outset. Before the unit exists, a UDL-informed teacher is asking: Why are we learning this? How will I present it in multiple ways? How will students engage with it? How will they show what they know? The three principles (engagement, representation, and expression) aren't checkboxes. They're what Katie Novak might call design orientations.

Many teachers and school systems treat UDL as a synonym for accommodation: extra time, modified texts, assistive technology. Those things matter, but they aren't UDL. UDL isn't retrofitting the curriculum for students who can't access it. It's designing the curriculum so the barriers don't exist in the first place.

Defining Differentiated Instruction (DI)

Carol Ann Tomlinson, who is arguably a leader in differentiation, has stated for decades that differentiation is a proactive mindset that a teacher brings to planning. Before the lesson begins, a differentiated teacher is asking: Who are my learners? What do I know about where they’re starting? What pathways and options can I build in so the learning reaches all of them? Multiple approaches to content, process, and product; not different destinations, but different routes to the same one.

Many teachers have accepted a version of differentiation that means reducing the task for students who struggle. Fewer requirements or something like that. Tomlinson calls this myth out directly: differentiation is more qualitative than quantitative. It isn’t giving some students less of the same assignment. It’s rethinking the nature of the assignment so it fits the learner while keeping the learning objectives intact.

Defining Scaffolding

Scaffolding is what happens once learners are in the room, and you can see what’s actually happening. Pauline Gibbons puts it precisely in her book Scaffolding Language, Scaffolding Learning. She writes:

In practice, scaffolding might look like a sentence frame that gives a student the language structure so their thinking can do the real work. It looks like a worked example that shows the process, not just the product. It looks like gradual release: I do, we do, you do. A think-aloud where a teacher makes their invisible reasoning visible. A guiding question that narrows the cognitive load just enough to get a student unstuck without removing the challenge. Notice what all of these have in common: none of them lower the intellectual demand. Second, with scaffolding, the aim is that students gain a level of independence to implement learning strategies on their own or with peers rather than relying on the teacher.

Bringing UDL, DI, and Scaffolding Together

Here’s a Claude Artifact (screenshot also below) of my understanding between UDL, Differentiated Instruction (DI), and Scaffolding as supported by several texts and AIs.

All three of these practices can occupy the same classroom, the same lesson, even the same moment. Consider a history teacher designing a unit on the civil rights movement. Before the unit begins, she thinks about how students will access primary sources, how they will engage with the material, and how they will show understanding through writing, discussion, or a visual product. That’s UDL doing its work at the design stage.

Within the unit, she notices some students need more time with the documents while others are ready to move into analysis. For students still wrestling with the sources, she designs a close-reading process. For students ready to push further, she moves them into a comparison and argument-building process. Different actions, same destination. That’s differentiation.

Then on a Tuesday, she opens a language scaffold bot she built in advance; one that knows the sentence starters, knows the argument structure, and knows its job is to practice with students until they can do it alone. A student who can’t connect evidence to a claim works through three or four cycles with the bot; it offers a starter, the student completes it, the bot pushes back gently, and the student tries again. By the end, the student has written the sentence. The bot didn’t write it. The teacher didn’t write it either, though she designed the whole condition that made it possible. The scaffold existed for six minutes. The student is ready to meet the standard and gains a sense of independence.

Over-Scaffolding and The Learning Pit

When a teacher manages every step of a lesson, students follow the path but never make sense of the terrain. Ready-made answers lead students to reuse solutions rather than build reasoning. Frey, Fisher, and Almarode put it plainly in How Scaffolding Works: without sufficient fading, students develop a dependency on the supports provided and fail to reach independence.

It's a little counterintuitive, but teachers need to allow students to sit in what James Nottingham calls "the learning pit"; that uncomfortable space of not yet knowing, which is where the real thinking happens. Tolerating that discomfort long enough for the thinking to happen isn't cruelty. It's the whole mechanism.

Monday Ready Resource: Prompt for Learning Pit Coach

When students get stuck, this bot helps them sit with the discomfort long enough to work through it rather than around it. A great addition to a “ask three before me” approach.

COPY AND PASTE INTO AN AI BOT FOR STUDENTS: You are a coach for students who are stuck and frustrated. Your first job is not to ask a question. It is to acknowledge what the student is feeling. Tell them directly that being stuck is not a sign that something has gone wrong; it is a sign that they are in the middle of real learning. Be warm and specific: the discomfort they feel right now is the learning pit, and every person who has ever learned something hard has felt exactly this. Only after that acknowledgment ask them one question: what is one small thing you could try right now, even if you are not sure it will work? If they say they don’t know, ask them to describe what they have already tried. If they say nothing, ask them to try one thing, anything, and come back and tell you what happened. Do not offer solutions. Do not explain the concept. Do not tell them what to try. Your job is to help the student stay in the pit long enough to find their own way out. Normalize the struggle. Trust the student.

Impactful scaffolding is responsive to the students in the classroom, their cultures, and their needs.

Studies across math, literacy, and language education confirm this: scaffolds built around one cognitive tradition can exclude learners who don’t share it. Erin Meyer’s research in The Culture Map helps explain the mechanism. Low-context cultures like the United States expect meaning to be spelled out explicitly; the task, the steps, the expected outcome, all stated directly upfront. High-context cultures like Japan, China, and much of the Arab world expect meaning to be inferred, relationships to be honored before instructions arrive, and the whole to be understood before the parts are named. A scaffold designed around low-context assumptions doesn’t just feel unfamiliar to a high-context learner. It can feel disrespectful, as if the teacher is being too direct or blunt.

And yet multilingual students don’t operate as fixed cultural types. Over a career working with international learners, I’ve seen students shift their communication norms depending on their language fluency, who else is in the group, and what they think is expected of them. As a supportive teacher, the best move is a genuine investment in knowing your students, paired with a process that keeps expectations the same while letting expression vary. The destination doesn’t change. Every student is working toward the same learning goal. What can look different is how they show the journey; one student writes a personal narrative, another builds a structured argument, a third talks it through before anything hits the page. The thinking and expectations beneath all of them are the same. This is where process-based learning has a real advantage: a framework like Think, then Generate, then Edit names categories of thinking rather than prescribing a single path or output. It doesn’t tell students precisely what to think, or how to think it, or how to show it.

Instead, it offers students a structure for thinking that they can internalize and repeat. And because each stage is named, teachers can be deliberate about when and how AI fits into the process, and when it doesn’t. The class can move through the same process together, while individual students work with different levels of support depending on where they are.

That same principle applies when AI enters the feedback conversation. A bot that opens by asking how a student prefers to receive feedback, before offering any observations at all, is doing something most fixed feedback rubrics never do: honoring the learner’s communication style before the content of the feedback even arrives.

Monday Ready Resource: Prompt for Feedback Coach

This bot guides students through seeking, making sense of, and acting on feedback using the Acquire, Analyze, Act process. The prompt below can help scaffold independence by gradually returning decision-making to the student at each stage. The goal isn't fluency with the bot; it's fluency with the process, practiced first with AI support and eventually carried into peer feedback cycles without it.

COPY AND PASTE INTO AN AI BOT FOR STUDENTS:You are a feedback coach working with a student through three stages. Before you begin, ask how they prefer to receive feedback; some want to hear what is working first, others want to get straight to what needs improving. Honor their preference throughout. In the Acquire stage, ask what they most want to learn from this feedback and what success looks like to them. Then ask them to share the feedback they received, or offer to give feedback yourself. In the Analyze stage, share one observation framed around their intention, then ask them to interpret it: what do they notice, what surprises them, what feels actionable? Do not tell them what to do. If they respond briefly, follow their lead and give them space. In the Act stage, ask them to name one concrete revision, make it, then reflect on what shifted and what they would seek feedback on next time. At every stage, the decisions stay with the student. Your job is to ask the question that helps them think one level deeper than they would alone.

Metacognition is one of the best scaffolds.

Metacognitive scaffolds that build in planning, monitoring, and reflecting produce stronger outcomes than those focused only on task completion. David Rock calls one planning move “prioritize prioritizing” in Your Brain at Work: deciding what matters most is itself a cognitive task that deserves deliberate attention before the work begins. When students know what steps to take and in what order, cognitive load drops. And as Frey, Fisher, and Almarode note in How Scaffolding Works, as students engage in deliberate practice with scaffolds and feedback, they develop habits that endure across time; what researchers call automaticity. The goal is for the process to become so familiar that students stop spending mental energy on understanding the instructions and start spending it where it belongs: on the thinking the task actually requires.

Linking metacognition back to the “learning pit”, when we practice metacognition before a learning task, we can anticipate the mindset, the strategies, and the places that we can get stuck, which makes it feel like it’s not a surprise, and students will know how to respond to said challenges as they happen. Without foresight into those realms, students will react impulsively to a frustrating situation, and the last thing they will want to hear is “that’s just part of learning! Get used to it!” In that situation, your learners are going to want an answer, and AI might be one click away to help them out. So again, the path to independence is to practice metacognition ahead of a challenging task and ask students to anticipate pitfalls and strategies to overcome them.

Monday Ready Resource: Article on Metacognition

I have an article about three AI-enhanced processes related to metacognition that connect to these ideas of planning, monitoring, and reflecting with AI that I mentioned above. Check it out below.

Monday Ready Resource: Prompt for Foresight Coach

This bot helps students plan before they begin using the Foresight stage of the Hindsight, Oversight, Foresight process. The goal isn't fluency with the bot; it's fluency with the planning moves, practiced first with AI support and eventually carried into any challenging task without it. Teachers can scaffold with a responsive approach by adjusting how students interact with the bot as they gain fluency; more guidance early, more independence later. Students who internalize the process start guiding their own metacognition in new contexts, perhaps without their AI coach and instead talking through obstacles with a peer, or working through the same questions in a journal before a task begins.

COPY AND PASTE INTO AN AI BOT FOR STUDENTS: You are a thinking coach helping a student prepare for a learning task. Do not help them complete the task. Begin by inviting the student to share anything about how they like to think through new tasks; some students like to see the whole picture first before breaking it into steps, others prefer to start with one concrete thing and build from there. Acknowledge their preference before moving forward. Then ask them three questions, one at a time: What is this task asking you to do? Where do you think you might get stuck, and why? What strategies do you already know that could help you work through those moments? After they answer all three, summarize their thinking back to them in a way that reflects how they described it, not just what they said. Ask if they want to adjust their plan before they begin. Always keep the decisions with the student.

Opportunities and pitfalls of scaffolding with AI

Scaffolding is a way of supporting students. The key to this form of support is that we intentionally plan its fade over time. So, it would be a misnomer to say that AI is a scaffold. Well, no. The way in which we deliberately use AI with a plan over time to support learning is a scaffold. The point being that it’s entirely in how AI is used, and the best way to get there is a well-designed, clear process with a plan.

Frey, Fisher, and Almarode describe distributed scaffolding in How Scaffolding Works as the in-the-moment support teachers provide while students are actively working; the nudges, questions, and hints that respond to where a learner actually is rather than what was planned in advance. They recommend a sequence: start with a question to check understanding, move to a prompt if that doesn’t unlock the thinking, then a cue, and only then a direct explanation as a last resort. That sequence is designed to keep the cognitive work with the student as long as possible.

This sequence lives within what researchers call the Gradual Release of Responsibility; the movement from explicit instruction (I Do) to guided practice (We Do) to independent work (You Do). Clark is clear that it’s not linear; teachers move back and forth between stages depending on what students actually need.

When a teacher tells a student, “AI is fine on this task,” the bot might help them skip the entire sequence and go straight to direct explanation. Seeking the path of least resistance is a human psychological trait. The move that works is specific: when you are brainstorming, you may use AI to push your thinking further. Not “AI is allowed.” AI supports this thinking move, in this way, at this stage. That is, teachers, name the thinking for each step of a process and how AI can support it.

If you haven’t seen my post titled “How To Design AI-Enhanced Processes”, it’s worth checking out. It covers the above ideas bolded above. In it are simple ways you can design a process that names the thinking, sets expectations, and considers what evidence you would find compelling to demonstrate student thinking.

The pitfalls follow the same logic as over-scaffolding more broadly. Frey, Fisher, and Almarode state in How Scaffolding Works that the most common error with graphic organizers is when filling out the organizer becomes the end goal: students turn it in, the lesson continues, and the opportunity to build schemas is forgotten. The same thing happens with AI. A student handed AI without a clear role or purpose probably won’t use it as a scaffold. More likely, they will use it as a completion machine. They finish without building the thinking that the task was meant to develop. And unlike a graphic organizer, AI is fast enough and fluent enough that the student may not even notice the thinking didn’t happen. The bots I shared in this article are designed to respond to each student in ways that honor their thinking and communication norms, asking questions before giving answers and holding back direct explanation as a last resort. But the bot isn’t the relationship. You are. Your role while students work with AI is to circulate, notice, and show genuine interest in what they are thinking. You are their biggest audience, and your curiosity about their ideas is what makes the process feel worth doing.

Conclusion

As a teacher, what you’re watching for is curiosity, critical thinking, grit, and metacognition. Those are the signals that the scaffold is working and that students are moving toward an independent, self-directed mindset.

Build a process that scaffolds agency with metacognitive routines. When we know those intentions and make them clear to our students, then we can invite AI into the learning. And when a student struggles, resist the urge to rescue them; instead, ask them if they anticipated this and what strategies they prepared. Phrases like “you know that it’s totally normal to be in the learning pit. Let’s take a minute to consider what options you have” go a long way toward establishing longer-term, sustained independence.

A dependent, transactional culture teaches kids something, too; it just teaches them that struggle means stop, that help means answer, and that learning ends at the final report card. So scaffold with intention.

AI Disclosure

In each article I write, I love to take different approaches in my process. Below, I have named my process and indicated when and how AI supported my writing. The feedback I have been getting from my readers is that these disclosures help them see how I model the practices I promote.

Time

I think that it’s important to share how many hours I spend writing articles because I want people to know that it is not necessarily about saving time. I still spend 8-10 hours writing each article. My process is different because AI is a part of the journey and I have access to a wider corpus of research as well. This particular article took me about 8.5 hours to complete across three days. The most time-consuming things were ensuring it captured accurate research and practices, ensuring I endorsed the ideas, and that the language was my own.

Process

To write this article, I spent three mornings waking up early, drinking some strong coffee, and going to work. I focused on getting the ideas down on paper, editing it once, then stepping away and editing it again with a fresh perspective. Since I often promote the approach of naming the thinking, I thought I would similarly share my steps here. Reflect, Write, Research (with AI), Edit (with AI), Edit (without AI by speaking with three fellow teachers), Record, Share.

Update: I came back a month after publishing this article initially. I learned new information about differentiation and updated this article. That’s the wonderful thing about an article-based approach: it is a representation of my thinking that can evolve over time!

Research with AI

In this step, I took my research questions and did a search in Consensus. While there is free access, I pay for it because I find that it greatly enhances my job as a coach, and I use it frequently. If I am going to recommend an instructional practice to a teacher, I want to know what the experts say. It’s a great way to get very specific questions answered with credible sources. On Consensus, I ran a report and summarized the relevant findings. I also included a couple of books I was familiar with and that were referenced throughout the article.

It’s a reminder that good teaching practices have a considerable body of research already out there. The question I like to ask when doing this sort of writing is, how does AI fit into the well-researched and impactful practices of teaching and learning, if at all?

Thank you for reading!

At a Glance

Feedback is only useful if you actually do something with it. That sounds obvious, but most students either avoid seeking it, collect it without making sense of it, or act on it without stopping to think about whether it moved them forward. Feedback requires trust. You have to believe the person or tool giving it has your growth in mind, and you have to be willing to be seen mid-process, before the work is polished. Acquire, Analyze, Act is designed to build both of those things: the habit of seeking feedback deliberately and the relational conditions that make honest critique possible. The process creates a repeatable loop; seek feedback, pause to construct meaning from what you find, then implement and reflect before starting again. AI can enter at any stage or none at all. Where it fits depends entirely on where students’ thinking needs a thought partner.

The Acquire, Analyze, Act process (below) is a flexible process when engaging in feedback, as it can incorporate AI at any stage; making it a different approach than the other processes shared up until this point.

Acquire

With a feedback-seeking mentality, consider what you want to learn.

• Set clear goals: Decide about what you want to learn, and define success using criteria that matter to you.

• Develop feedback-seeking behavior: Expect revision, regulate emotions, and build trusting environments where critique is welcome.

• Seek diverse perspectives: Collect timely feedback from multiple sources (AI, peers, teachers, self).

• Time it strategically: Seek input early and often, when you can still implement meaningful changes.

How might AI be used in this step of the process? Create a bot in which the learning objectives are entered by the student or teacher. The students share their work and discuss the desired outcome.

Analyze

Intentionally pause and construct knowledge by examining the data

• Interpret: Look for patterns, agreements, and contradictions across feedback.

• Unpack: Pursue “why?” and “show me?” until the data is crystal-clear.

• Identify takeaways: Select the most impactful actionable insights.

How might AI be used in this step? Upload a summary of feedback, annotated work, etc., and ask for a summary of and actionable next steps to a large language model like Chat GPT. Have a conversation with the bot about what the next steps might be.

Act

Take time to implement your takeaways

• Iterate promptly: Apply insights while they’re fresh and check revisions against original objectives.

• Repeat as necessary: Return to seeking feedback and continue the cycle until your goals are achieved.

• Reflect: Consider how the feedback process helped your thinking and encouraged growth. What would be helpful to seek feedback on during the next cycle?

How could AI be used in this step? Students might use AI to vibe code (when we code with the support of AI) and ask it to support our vision based on the feedback we received. There are many possibilities!

At a Glance

This is one of my most popular processes that can be used quite widely. I have found it to be the most helpful for classes practicing creative thinking which I define below in a graphic titled “Traits of Creative Learners Who Use AI”. The core of this process is that we start and stop with human thought, creativity, and analysis, and when we use AI, we have an iterative dialogue, rather than simply taking the first output with little consideration.

The Think, Generate, Edit process (below) has specific verb-based actions that you or your students can use when being creative. The intention is that we begin and end with our own efforts, and in the middle, we collaborate with AI, rather than letting it do the work for us.

General verbs were used in this process so that you can define them for your own creative context. For example, the word “think” could be made more specific by saying, “think creatively by brainstorming a range of possible story ideas.” The second step could be “generate by expanding on one idea, using a text-based AI to explore alternative openings, settings, or tones”, or “edit by curating and refining the strongest pieces, weaving them into a coherent draft that reflects your own vision.”

Think

For the purposes of this process, the word “think” is intentionally vague. I wanted to use a term that could encompass both creative and critical thinking.

Before engaging with AI, we should develop high-quality content that reflects our intent. This might take various forms: crafting an essay that captures our distinctive voice and perspective, composing a song that expresses our musical ideas, or creating a photograph that demonstrates our understanding of lighting, composition, color, and visual storytelling. It could even be in the form of a well-thought-out prompt. Whatever the thinking might be, it serves as the foundation for the AI interactions in Step 2.

Thinking with AI could look like creating a pencil sketch, a song, a prompt, having a chat, a photograph you took, a whiteboard mind map, and so much more. With multi-modal AI models, we can input just about anything that expresses our thoughts.

Generate

Output involves an iterative dialogue with AI tools. Through careful prompting, refinement of inputs, experimentation with different AI models, and continuous adjustment, we work toward obtaining results that complement and enhance our creative vision or intent. This stage requires patience and strategic interaction to achieve the desired outcome. Furthermore, you are engaging in both creative and critical thinking throughout this step: what did you create? Does it meet your needs? If not, how can you adjust and do it again to get precisely what you want to express your vision?

Edit

This step is crucial because it’s where we critically examine AI output and make it our own. Through human judgment, expertise, and creativity, we transform what the AI generates into something that aligns with our original intent.

Although it appears as the third step, editing is rarely the end of the process. Creative work is iterative, and we often move back and forth between thinking, generating, and editing as ideas evolve.

At this stage, I intentionally recommend stepping away from AI. Editing is our opportunity to slow down, apply analytical and creative thinking, and ensure the final work reflects a distinctly human voice.

Last week, I was standing in a 7-Eleven in Nara, Japan on spring break, and before setting off to explore the charming city, I stopped to buy an onigiri rice ball as a snack. While checking out of the 7-11, I remembered something from the time when I lived in Japan. My friend Soichiro taught me how to open onigiri about twenty years ago by following the numbers on the packaging: three tabs, a folded plastic wrap that keeps the seaweed crispy and separate from the rice until the exact moment you want them together. Precise folds, purposeful sequence, color-coding— to me, it was the kind of design that seemed to draw upon the wisdom of origami.

Check out the video below of me showing the packaging of an onigiri and how opening it is easy and leaves the seaweed dry and crunchy. Fun fact: 7-11 wraps their packaging in bioplastics!

Actually, Soichiro was not there the first time I tried to open one on my own. I just started pulling at the plastic like I was unwrapping a granola bar. I tore straight through the seaweed, the rice went everywhere, and I ate a slightly soggy, structurally compromised snack standing outside a convenience store, feeling very foreign. The packaging already had the answer, though; three numbered tabs, right there on the wrapper. The design was not the problem. I just didn’t stop to read it.

Over the remainder of my trip, I kept noticing the same well-designed logic everywhere, from vending machines to train exit gates to conveyor-belt sushi restaurants. One of my favorite designs was a paper cup dispenser with a single button to release exactly one cup from a locked stack. I watched a tourist wrestle with that type of machine for thirty seconds before noticing the button. Back when I lived in Japan, I learned that when I struggled with something like a paper cup dispenser, the right response was to self-correct. That is, if something is difficult to open, use, or do, you’re probably doing it wrong. In Japan, the user experience is often carefully planned and meant to be easy.

I came home thinking about teaching and learning, and I kept thinking: what if we applied the same logic to classroom instructions? Much like a wrapper with instructions, classroom instructions should be easy. The task should be where the energy is put. Students’ effort belongs to the thinking, not to decoding what you want them to do. In other words, opening the onigiri was not the point. Eating a delicious snack was. The packaging exists to serve the experience, and the best packaging gets out of the way quickly. Classroom instructions work the same way in that they are the vehicle for learning, and not the purpose or when learning happens.

Picture a high school student with four classes, each coming with lengthy instructions and teachers who carefully cover every edge case before anyone touches anything. By the time a student opens a task on their computer, they are more glazed over than a honey-baked ham! And because we live in an age in which everyone is using AI, they’ve probably got their favorite model running in the background of their laptops. Once they reach the point that the instructions become overwhelming, the internal monologue becomes: I honestly couldn't care less. I'm exhausted. I just want to get through this. This classroom and day-to-day experience sets kids up to have a mentality that is vulnerable to AI misuse. Kids who feel less engaged and disinterested will want to complete tasks quickly, and AI can provide a shortcut.

If your instructions lose them from the get-go, you’re heading in the direction of compliant task completion. Too much teacher talk that muddies the instructions might indirectly push them toward feeling overwhelmed and toward a desire to cognitively offload the task as efficiently as possible.

My suggestion is this: get into the intellectually engaging, stimulating process of active learning in class. The better you can design your instructions to be short, verb-based, and clear, the better. If you are noticing friction with instructions, processes, or any element, that difficulty is highly informative and can help us to adjust.

So in other words: difficulty is data.

The Look on Their Faces

A quick clarification before I go further. Direct teaching is a powerful tool (see Hattie’s work). There is absolutely a time to stand at the front of the room and teach. This article is not about that moment. This article is about when you ask students to do something, and you are explaining how to engage (e.g., create, discover, reflect, collaborate, analyze, build). The task is meant to generate learning, and before any of that can happen, you have to explain what to do. From my experience as a teacher and coach, fifteen minutes or less with an exemplar is the limit.

When teachers overexplain instructions, it leads to a kind of glazed-over, fading anticipation mixed with compliance. It’s funny too, kids will avoid asking questions because they just want to get on with it, even though they actually have many things they want to ask you, they bide their time and plan to ask a classmate what they are actually supposed to do.

Myth: good instruction means frontloading every common misconception and pitfall before students have touched the work. To be clear, anticipating roadblocks is good design; that is what Universal Design for Learning asks us to do. But there is a difference between designing for barriers and narrating all of them upfront before students have had a chance to think. When teachers over-explain every obstacle in advance, they usurp the learning; students never have to construct cause and effect for themselves because the teacher already did it for them. They arrive at the work with a head full of caveats and nothing left to figure out. That is not so different from handing a task to AI in that the thinking gets outsourced before it ever begins. Just as we don’t want AI to do the work for students, we also don’t want teachers to do the work for them either.

I used to be the over-explaining guy: I’d hover while students work, point at their screens, announce new pitfalls I just remembered or noticed, and announce that there are thirteen minutes left. I would not necessarily call that a rich thinking environment; you know what kids are thinking in that situation? I’m going to just get through this block so I can go home and do it on my own, and I’ll just ask AI and my friends if I get stuck.

Could you imagine if 7-Eleven sold onigiri that required 27 steps to open, and a lengthy training video that walks you through every possible way it could go wrong, and then you are given 13 minutes to do it, while in the back of your mind you know that you have a really important train to catch at the station? You would be exhausted, uninterested in the snack, stressed, and looking forward to the whole thing being over.

If we are explaining the instructions to an activity and the students have their heads down, that’s data. It is the equivalent of struggling with an onigiri wrapper. It does not mean your students are necessarily unprepared. It could mean your instructions have friction in them, or the students are just not paying attention due to distraction, confusion, or feeling overwhelmed. Every minute a student spends decoding your instructions is a minute they are not spending on the actual thinking you designed the task around. That thinking, the brainstorming, the analyzing, the revising, the reflecting, is where the learning happens.

Teachers are designers who are constantly testing their products and empathizing with their clients. So with that design thinking mentality, when students look lost before the learning starts, we can think of this as an observation in which we ask ourselves: what did I build here? What can I subtract? How can I activate thinking and step out of the way? How can I provide just-in-time feedback?

Monday-Ready Moves

Here’s a list of a few strategies that I have seen work as a teacher and coach. They directly support process-based learning in that a strong process can actually serve as clear instructions that do not necessarily require lengthy explanation.

1. Limit teacher talk. Read your instructions once and keep the total instructions to 15 minutes or less. The shorter your instructions, the more energy your students will have. If you are still talking after 15 minutes, something needs to come out, or additional instructions can happen later in the same lesson. Again, this is not for direct teaching in which essential information has to be taught; I’m talking about the instructions for an activity.

In terms of designing a slide, make the words large and easy to read from across the room. Don’t write all the instructions, just the main points so they can recall what they’re supposed to do.

2. Lead with an exemplar. Show before you explain a model paragraph, sample sketch, before-and-after comparison, etc. When students can see the destination, your words serve as confirmation as they build theories about the task and its outcomes, rather than as orientation.

3. Use verbs to name the thinking. Replace vague nouns with precise action verbs. Not “work on your essay” but argue, support, challenge, revise. Not “think about the data” but interpret, compare, decide. Verbs tell students what their brains are supposed to be doing. They also support clear expectations about where AI can or cannot do the move for them (#4 below). For more independent students, you can also ask them to engage in metacognition before starting by considering which steps in the process would be most strategic for meeting the learning objective, then, as they are ready, proceed with their own.

4. Name the AI expectation for each step. For every thinking move, students need one clear statement: what do I do, and what does AI do here? For example, “AI will give you counterarguments, debate it, then record your key findings in your process journal.” Another example could look like, “No AI on this step; this is your thinking.” Vague AI expectations can invite interpretation. A single sentence per move removes the guesswork and keeps the cognitive effort where you want it: on the learning, not on figuring out the rules. For older and more independent students, you can also co-design AI agreements together.

Students also tend to like when teachers allow AI on a step in a process that they create a purpose-focused bot for them to use (e.g. on School AI, Flint, etc.) That way they feel less anxious about misusing AI on the task. Students all have access to AI and will likely use it at some stage, but you can make doing the right thing easy.

5. Tell them “the why of the how”. Once students know what they are doing and how AI does or does not support each step, add one more sentence: why this particular approach? “We are using journals here because writing slowly pulls thinking out of your head and away from the distraction of your screen.” Another example might be: “We are using AI to role-play as an audience member so you can practice your speech and feel confident before you perform it for real.” Using that language is a direct signal a teacher can send: I thought carefully about how you learn, and I chose this because I want you to succeed. That message contributes to belonging, trust, and independence because it helps students see a process as something they can actually use again on their own.

Conclusion

I am not saying instructions should be dumbed down. I am saying they should be carefully designed, succinct, and clear to help students access the thinking that leads to deeper understanding. Dumbing down removes the challenge and presupposes students’ incapability— what you might say is setting low expectations. Focusing our language during instructions removes the confusion so the students can engage in the process and put effort where it matters.

Students should struggle with the counterargument, wrestle with the revision, sit with the discomfort of a claim that does not quite hold up yet. That productive struggle is where growth happens, and it is worth protecting. Let’s help students put energy into that thinking and not want to turn to AI for task completion.

Look for signs that students are thinking. They are writing, discussing, touching their faces, drawing, reading, etc. If they move immediately into the thinking, you have built something that facilitates their thinking. If they look low in energy, slumped down, or frustrated, change something.

The second time I opened an onigiri, even with Soichiro showing me how, I still tore the seaweed a little, and that moment of friction is exactly what made the experience stick. I learned from it and gained independence from Soichiro as my teacher. I never tore it again and actually went on to show my friends how amazing it was to eat.

When it came to that onigiri, though, the packaging was never the problem. The instructions were printed right there with three numbered tabs. When I tore the seaweed, it was not because the design failed me; it was because I jumped straight in without reading. The moment I paused, or when I had an example to follow (Soichiro modeling it), it worked. That is the other half of the teacher’s job. Write instructions that are clear, yes. But then create a pause, ask about their clarity, read them aloud together, point to the exemplar, ask students what they will do first, second, third. Give students a moment to actually look before anyone touches anything. The packaging can be perfect and still get ignored if nobody stops to read it first.

When they struggle with the task itself, that is not a problem; in fact, that is the point! Struggling with a counterargument, sitting with a claim that does not quite hold up, wrestling with a revision that keeps slipping, that is productive. That difficulty is data too. It is just pointing at the learning instead of the design.

AI Disclosure

This article was started with Claude 4.6. It started with me dictating long and disjointed ideas on a train leaving Osaka, using voice-to-text as a way to capture a loose set of noticings and half-formed ideas\. I shared the transcript with Claude and used the conversation to brainstorm, push back on my own thinking, and gradually move from a vague noticing into something more structured.

Then I took a two-hour Shinkansen up to Tokyo and looked at misty mountains and patches of cherry blossoms along the way. My favorite German electronic music, Apparat, on headphones. I had time to think, write, and look out the window at nothing in particular, which turns out to be one of the better conditions for getting ideas to settle into something real.

I want to be transparent about what that human-AI collaboration looked like, because I think it matters. The ideas in this article are mine; the experiences are mine; the framing is mine. Claude helped me organize, sharpen, and refine them. There were moments it wanted to take things in a different direction, toward UX frameworks I did not need, or toward sensory details that sounded vivid but making my writing verbose.

I share this because I believe we should normalize transparent disclosure of AI in writing, especially those of us asking students to do the same. If you are not disclosing, you could be modeling secrecy.

I wonder if I were to name my process, it might look something like this:

1. Speak. Use Claude to organize your verbal ideas. Think outloud and share what you are currently thinking, then the AI can help you to organize them into a narrative outline.

2. Develop. Take the outline and write your ideas out. Do this step without AI.

3. Edit. Take your first draft and show it to Claude. Ask it what it thinks about your flow and connection of ideas.

4. Revise. Independently on a bullet train in Japan, re-read your draft a few days later with a fresh perspective and consider your draft. Edit it using your expertise as a teacher and ensure the article has clarity for your given audience of educators.

5. Share. Schedule your article to go out on Monday morning to share with your community.

After the content was written, I went back to add extras to make it more engaging like pictures for each section and a video of me opening the onigiri. The video was me in the streets of Nara actually opening a tuna and mayo rice ball, but the audio was enhanced using Adobe’s Voice Enhancer. The pictures throughout the article were cited with the model and process I used in their creation; all were generated using ChatGPT or Gemini.

I also want to state for the record that the en and em dashes used in this article were all me! I’m reclaiming them!

Thank you so much for reading this article. If you found it helpful or enjoyed its content, please hit the like button, or go one step further and send it to someone who you think might like it.

Share

If you want to try building your own process, I have a free tool at aiep.lovable.app that walks you through it step by step.

Students lose most of what happens in a school day because sense-making never occurs. Documentation, or “breadcrumbs,” gives students a visible trail of their own thinking so reflection becomes a real conversation with their process rather than a shrug at the dinner table. When AI is in the room, that trail matters even more; it forces the question “who did the work here?” Done right, breadcrumbs serve the learner, not the gradebook.

Another Dinner

Have you ever had dinner with a teenager (maybe you were this teenager) and asked, “What did you learn at school today?”

You get the classic “nothing,” delivered with a shrug that somehow manages to be both dismissive and existentially exhausted. You try again, “But you were there for the whole day. Nothing at all? C’mon, how was your Math class? What’d you do in that class?” Another shrug. The teen sets their fork down for emphasis of exasperation.

The thing is, a lot of things happened.

Our eyes, ears, noses, mouths, and fingers are constantly feeding us sensory input; that’s how we gather information from the world and make meaning from it. The problem is that a single school day throws an enormous volume of input at students all at once.

Class after class, concept after concept, and somewhere in that chaos, teachers, students, and parents assume that something meaningful happened. It reminds me of a photography technique called “spray and pray”: take as many photos as you can without being deliberate and hope that you magically get one usable photo.

Teen angst aside, what if students aren’t being entirely evasive when they say “nothing”? They have nothing in the forefront of their minds because sense-making didn’t really happen. Without time to consolidate what happened into meaningful takeaways, insights, and actionable plans, and without a record of the process to reflect on, the day’s data becomes noise. That noise gets worse with AI as a guest in the room; the temptation to skip the hard parts of learning is certainly made more attractive when AI can just do it for you.

Meaning Making

There are many ways that educators can support students to make sense of all the things that happen in their time at school. Structured thinking routines, the kind that ask students to reflect, notice, compare, and draw conclusions, are one approach to the noise problem. Furthermore, when those routines are paired with visible documentation of a thought process, students can actually watch their own thinking develop over time, notice how one decision led to the next, and see their process as something that grew rather than something that just happened to them.

As a teacher, I have seen documentation work again and again, but there’s a pitfall that I often write about: this only works when documentation isn’t treated as compliance because it stops serving thinking and starts serving control. Teachers who take a compliance approach are usually the same ones who say documenting learning is extra work, because for them, it is. They’re checking every step rather than coaching the thinking by sharing with students why documentation supports them to be successful learners and thinkers. They love to hear that!

Another key point to share is that teachers often assume that we know how to structure reflection. This belief comes from the fact that we often see the value in reflection, and we’re also taught in teachers’ college to do it, but in practice, it is not always implemented with impact.

What if we were deliberate in our approaches in ways that help students notice their learning a little more clearly? Structured reflection for students comes with documentation of learning as a form of data. When students have a visible trail of their own thinking, reflection becomes a conversation with their own thought processes and evidence of work. Done well, it takes 30-45 minutes with discussion and meaning-making (rather than a rushed 5 minutes at the end of class when they’ve already packed their bags). When students document their thinking, they can be invited by their teacher to see what they actually did, notice what changed, articulate why, and extract learning from the messy learning experience. But more importantly, they have something to point to and discuss as a third point in their conversations with others. Reflections are powerful when there is an audience who is interested and curious to know more.

A process for a structured reflection can be simple! Students could journal to prepare to talk with a peer, then have a turn and talk to share, then return to the journal to consider why things happened, what things they’re learning, and what their next steps could be. If AI were a part of the process, it might be a Socratic partner in the journaling stage to help children think about their work up until that stage. So in that way, AI becomes a practice and supports the meaningful relationship between people, which is one of the important things we get from our experience in school. Relationships.

In this singular, simple, and powerful process, we can foster consolidation and understanding, rather than what could be meaningless action focused on completion of tasks and exasperated dinner conversations with dramatic fork-laying. Through metacognitive routines, students go from being exposed to information to carrying it forward as new learning.

I really want to make it clear that documentation serves the learner first. That doesn’t change when AI enters the room as a guest collaborator on any thinking students are doing. When a student works with AI, the breadcrumb trail of their thinking illustrates what they did; they can see it, reflect on their process at a metacognitive level, and be deliberate in their human-AI collaboration and ask, "Who did the work here?” This is an important question we ask in an AI-enhanced classroom during our metacognitive consolidations. I will share more questions in the Monday-Ready Resources toward the end of the article that you can use to help your students reflect on their thinking.

What Do The Experts Say?

I was reading The Power of Making Thinking Visible by Ron Ritchhart and Mark Church, and there’s a section that I enjoyed. Their Project Zero colleagues define documentation as “the practice of observing, recording, interpreting, and sharing, through a variety of media, the processes and products of teaching and learning in order to deepen learning” (Given et al. 2010). I highlighted that quotation and more in the section below because it’s just that good. Give it a read!

What I love about this is that documentation serves multiple purposes: observing, recording, interpreting, and sharing, all aimed at deepening learning. No mention of policing or controlling students. When students observe their own process, record what they notice, interpret what shifted, and share their thinking with others, that is metacognition in action! Documentation is the mechanism that makes it possible.

“But Isn’t This Just More Work?”

If you’re asking about the workload for you as a teacher, you get to choose how much of it you read. Students can review their own thinking and peers can respond to each other. Teachers don’t have to be the center of all knowledge, learning, and effort in the room! Isn’t that freeing to remember? The teacher who says “this feels like more grading” is picturing a system where everything funnels back to them; let your students do the majority of the effortful thinking!

Second, documentation doesn’t have to be complicated. It could be a reflection, sketch, or screenshot of an AI chat with a two-sentence takeaway. It can be as complex as the student needs or a record of what they did in that step of their thought process. It could be as simple as a revision note explaining why a student made the choice they made or as complex as a research plan. You know your students and you know the learning objectives. The point is that we record students’ thinking to help them see how it unfolds over time.

Third, the way documentation is framed can have an impact on our mindset around it. As a coach, when I use the word “breadcrumbs” with teachers instead of “documentation,” they tend to like it. There’s something about the small size, and the Hansel and Gretel allusion to leaving a trail so you can see your way that resonates in a way that “documentation” never quite does. Try it with your students and see how it goes. Try reading a part of Hansel and Gretel with your students, too, even if they’re in high school. It’ll be a fun invitation to documenting learning that they won’t see coming.

Fourth, Not every breadcrumb needs a grade, and that’s usually what teachers actually mean when they say documentation is more work; they’re picturing themselves assessing every artifact, every reflection, every sketch. A single breadcrumb doesn’t always reveal much on its own, but a trail of them over time reveals something to the student and teachers (as coaches, not police) that a final product never could in isolation; it’s the basis of holistic assessment, where the whole collection tells a richer story than any single piece of evidence, or worse, none at all. And while we’re here: who decided the product has to be final? Sharing, reflecting, and iterating are legitimate steps after a “final” essay is “done”. The breadcrumbs are what make that continuation possible; they’re the trail and the collection together, the picture of what actually happened and how a student’s thinking evolved alongside AI over the course of a larger process.

So really, if you find yourself saying that documenting learning is more work for you, you might be doing more work than the students. The fabulous thing is that the locus of control sits squarely with you! You can choose how much of the documentation you read, what student-peers give feedback on, and what students use to self-reflect. You need not worry about carrying this on your shoulders, my friends!

One trap worth naming: students who are highly motivated by grades will sometimes document their thinking for the wrong reasons. If a teacher responds to “is this graded?” by grading every step, the breadcrumbs become compliance artifacts and the whole purpose can be subverted. A more powerful move is to teach the why instead. Something like: “We’re collecting our thinking in this journal because it’s one of the most powerful ways to see how your own learning actually works. I’ll ask you to reflect and notice what’s changing, and to plan your next steps. This is what deliberate learners do, and I’ll be with you through the process.” That framing invites students into the science of learning rather than the transaction of grading.

What Can Breadcrumbs Actually Look Like?

There’s no single right answer for the best ways to document thinking. You’ve got to use your professional judgement and awareness of your students’ needs, abilities, cultures, interests, and background knowledge. So think of the below ideas as a menu of possibilities. Pick what fits your context as the teacher, or if your learners are ready, give them choices.

Process Journal 1: Google Slides or PowerPoint

This is a format I keep coming back to because of its flexibility. Slides chunk thinking automatically. Pages duplicate easily. Teachers can co-design simple templates with students for frequent moves like reflections, research notes, journal entries, and prototype sketches. It’s super easy to take a photo and drop it into a slide. They’re also made for presenting. During a class share-out, a student’s process journal can go full screen. It also works beautifully for student-led conferences, where the process journal becomes the artifact that a student walks their family through. Free, easy, and high impact.

If students use AI in their thinking process, screenshots and links can easily be placed into this format.

Process Journal 2: Google Docs or OneNote with Tabs

If you want to combine the idea of PowerPoint and Word into a one-stop shop, both Google Docs and Microsoft OneNote are pretty good. They both have tabs to show sections, time periods, or even stages of their process. A second thing that can be used to support student thinking is revision history, which shows the timeline of thinking. This is available in just about all Microsoft and Google tools, although I have not seen it in OneNote.

If students use AI in their thinking process, similar to the above, screenshots and links can easily be placed into this format.

Process Journal 3: Paper Notebooks

This option is less shareable, so if that’s your goal, you might want to try Slides. That being said, paper journals are often more focused. When students step away from screens, something magical in their creativity and depth of thinking happens. Students frequently say it actually feels good to put the tech down for a bit. There’s something about pen and paper that resists the temptation to edit and focus on expressing ideas slowly.

If students use AI in their thinking process, it might take a little bit of creativity and trust to show how they worked with it. Perhaps printing and pasting screenshots like a scrapbook and writing their thoughts.

Process Journal 4: Evidence Folders

This format is helpful when there’s a wide range of artifacts that are disparate in nature. You might ask the students to follow clear naming conventions and organize their work in a way that is easy to follow. The folder could include screenshots, audio recordings, movies, links, PDFs they read, and anything else they find worth saving.

If students use AI as a part of their process, they can easily include links to entire chats, screenshots of conversations, or Word files with both of the aforementioned things.

Conclusion

Teens will likely always be exhausted at dinner when you ask what they did at school today. That’s not changing because teens are teens, and we love them just as they are.

But exhausted-and-filled-with-noise is different from exhausted-and-clear. A student who practiced metacognition with their artifacts through structured conversations and had time to do so in class before more content was thrown their way will have a much clearer sense of direction. It’s the power of consolidation.

I’m worried that AI could add to the noise and the desire to complete more tasks. Let’s be more productive, do the job of 10 people, and complete more tasks without actually having gained anything.

I would love to teach our way out of a dystopian, AI-reliant future by empowering thoughtful children with the practice of reflection through documentation. When AI is a guest collaborator in the room, that kind of awareness has to be practiced.

Teens might not volunteer everything they learned. Parents might still need to ask twice. The fork might still go down. But every now and then, a student reaches back into their day and finds something worth talking about; something they built, something they noticed, something that was actually theirs to share.

Monday-Ready Question Bank

When asking students to practice metacognition with the documented breadcrumbs, you might ask these sorts of questions. I have framed them almost as “before and after” with commonly asked questions reframed.

• Instead of “What went well?” ask “What shifted in your thinking, and what caused that shift?”

• Instead of “What would you do differently next time?” ask “What’s the most important decision you made in this process, and would you make it again?”

• Instead of “What was challenging?” ask “What challenged you, and did you push through it or find a way around it?”

• Instead of “What did you learn?” ask “What do you think now that you didn’t think at the start, and what changed your mind?”

• Instead of “Did the AI help you?” ask “Who did the work here? Where was the thinking yours, and where did you hand it off?”

Related Reading

Take a look at the example below from David Gran’s article for InThinking in which he created a film with his thoughts about AI. The article is an exploration of ideas, but it’s also a process journal with several interesting artifacts. He has screen recordings, animated GIFs, pictures, and other forms of documentation that could inspire your work with your students. There is so much transparency in his process!

Related Podcasts

Below are a few podcast articles that I thought you might also enjoy listening to, related to the idea of documenting learning, metacognition, and assessment.

AI Disclosure

This article has been all about breadcrumb trails of thinking, and I have shared many possible ideas. This disclosure statement is another example of what a breadcrumb could look like in the form of a disclosure statement.

This article was developed with AI assistance, including content drawn from the Metacognition chapter of my book, AI-Enhanced Processes (Second Edition). The direction, framing, examples, and voice reflect my judgment throughout.

I used Claude 4.6 Sonnet as the primary collaborator on this article. I found Projects to be helpful because it was aware of my book, my style, and purpose. That helped me to stay on track and to have a critical friend in the process.

Images were made primarily with Google Gemini, although at times I used ChatGPT to get the image started. Gemini was definitely fast, but would sometimes create images with inconsistencies that needed to be adjusted. I enjoyed Gemini’s annotation features to be able to show the AI where I needed to change a picture.

This article took me approximately 9-10 hours to write, and the editing process was completed while on an airplane ride, while offline and flying to Osaka. AI supported the drafting process; the thinking is mine.

The Generator collects a few types of input. First, the thinking moves you want to center; the verbs that matter in your classroom. Second, your class context; subject, task type, grade level, and where you are in a unit. Third, your learners; cultural context, language needs, scaffolding considerations, anything that shapes how your students engage. Fourth, the ways students can make their thinking visible through documentation or what is referred to as “breadcrumbs” (i.e. trails of thinking).

The site was vibecoded with Lovable.dev and iterated upon over the last couple years. It runs on DeepSeek by taking your input and sending it via what’s known as an
”API call” and returns a custom process and is completely free to you (I pay for each API a fraction of a penny). It takes about ten minutes to create a process. You might think of it as scaffolding to get you to where you want to be with process creation.

It’s important to note that this is a sketchpad, not a finished process you can copy and paste into your classes. This tool gives you a reasonable draft and gets you and your students started. You, the wise user of AI, still bring the judgment. That part does not get outsourced. So please practice Think, Generate, Edit whenver using it. Think by considering who your learners are, what you want to accomplish, and other important details. Generate by following the website’s prompts. Edit by judging the output, editing your notes, and regenerating until it’s precisely what you and your learners need.

How to Use It

Name your thinking moves, AI’s role, and Where AI is Placed

The generator asks you to identify the type of thinking you want students to practice. This is the heart of an AI-enhanced process; you have to name the thinking before you can design around it. Consider being specific about AI’s role and where it should appear within your process. Do you want a coach-like AI that is used at the beginning of the process? Those are design decisions that you can play with; don’t worry if it doesn’t turn out the way you had planned, you can try different options and regenerate the process at any time.

Add your class context.

Tell the tool what you are actually teaching. The task, the subject, the grade, where you are in the unit. This is what makes the output useful instead of generic. “High school English, argument essay, mid-unit revision” produces something very different from “middle school science, claims and evidence.” The more concrete you are here, the more helpful the generated process becomes.

Add whatever shapes how your students engage: home languages, prior knowledge gaps, cultural contexts that matter for the task, scaffolding needs. The generator uses this to make the process fit your actual humans, not a hypothetical class of compliant, monolingual, grade-level readers.

Generate.

The tool produces mostly three-verb processes that should be thought of as a first draft and iterated upon. No AI will get precisely what you and your learners need the first time. It will also produce a detailed prompt built around your inputs that you can feed into a bot so that it support learning precisely the way you want within a process. Both the process and the prompt are yours to use, modify, and run.

Didn’t get what you want? Regenerate!

The site was designed as a single page intentionally to help users create processes by refining their prompts and tweaking their results until it matches their needs.

Download as a Word File and Build AI Bots.

You’ll notice that after you generate, at the bottom of the page is a download button that gives you access to a Word file with all your details, learning objectives, the process, and a prompt you can put into your school’s AI platform (e.g. Flint, Magic School, SchoolAI, etc.) The bot will know precisely its role within your process and your learning objectives. Below is a preview of the output you’ll get from the Generator.

Teachers: Taking the Prompt Into Your Classroom

The generated prompt is designed to drop directly into a purpose-built bot. Paste it into SchoolAI, Flint, Magic School, or whatever platform your school uses. Once it is loaded, the bot knows your process; it can coach to it rather than around it.

Here is what that looks like in practice. Say your thinking move is brainstorming and you want students to generate their own ideas before AI enters the room. The prompt tells the bot to ask Socratic questions that push student thinking; it will not generate ideas for them. Students have to bring something first. The bot responds to what they bring. That is the guest collaborator framing in action: you set the invitation terms before the guest arrives, and the guest respects the boundaries of the process.

Students: Designing Your Own Process

If you’re a teacher and you’re reading this, feel free to share the site with your students when thare are ready to take charge of their own process design. What we ultimately want to see is that students show a level of independence and self-directedness in their learning. Afterall, as adults, they probably will not have someone creating scaffolded learning experiences to support the meaningful use of AI to support and not replace thinking. It is therefore arguable that somewhere in secondary school and university that students will likely be ready for increasing amounts of independence in their thought processes that leverage AI.

Please feel free to encourage students to intentionally consider when and how they use AI in their work. Asking them to use this site is one way that they might consider their thinking moves with your gradually released support, feedback, and coaching.

What It Does Not Do

The tool does not know your school’s AI policy.

It does not know which platforms are approved, what your administration has communicated to families, or what your district’s guidelines say. It also does not know your values or how you express them through actions. That context is yours to bring, so please share it in the appropriate fields when generating your process to make the output contextualized for your students. That being said, this tool is more for thought process generation and will incorporate additional details where possible. The school’s culture, values, policies, etc. are likely going to appear around how this process is implemented.

It will not know your students the way you do.

The draft will be a starting point, not a finished design. The first version will probably be a little messy. That is fine. Messy first drafts are how good processes get built; you run it, notice what doesn’t quite work, and iterate. Remember, nobody is an expert on AI and you should give yourself permission as a lifelong learner to experiment.

Do not expect automaticity on the first run.

Students need to practice a process repeatedly before it becomes internalized; a new and novel AI-enhanced routine does not become second nature after one practice. Plan to run it, reflect on it, and revise it with them! Ask your students for feedback: how could we make this process even better next time? The tool makes the first draft fast. The repeated practice is still your role as a teacher to implement in class repeatedly. Think of it as a strategy on how we use AI effectively to support our thinking; it’s a strategy that takes practice. In that way, the process could be displayed in the room on the walls to remind everyone of the strategies that can be leveraged when needed. A few, high impact strategies that are practiced throughout the year is a great approach.

Invitation

Want to give it a try? Make your own process! Help students make their own! Let’s share your story. Reach out to me via the “About” page or LinkedIn and let’s share your story about the process used to support student learning.

Not sure where to start? You do not need to redesign your entire curriculum. Pick one task you teach repeatedly; brainstorming, outlining, peer feedback, revision, reflection. Something you already have momentum with. That is your best candidate.

Go to aiep.lovable.app and follow the steps. Take the output, tweak it until it sounds like your classroom, and run it once this week.

Planning a process used to mean starting from nothing. Now it means starting from a reasonable draft and making it yours. That is not outsourcing the thinking. It is protecting time for the thinking that actually matters; the judgment calls, the coaching moves, the relationships that no synthetic guest can replicate.

Name the thinking. Set the expectations. Make the learning visible. The tool helps you get there faster.

Here’s a free poster you can use with students to build language around your shared AI values.

Shared AI Values (A3)

7.14MB ∙ PDF file

Download

Download

TAPE, Two Years Later

In 2023, I stood in a room in Hong Kong at Apple’s Future of Learning HK conference and scribbled down four words that would eventually shape everything I built after it.

Holly Clark was facilitating. The conversation was electric. And somewhere in that room, TAPE was born: Transparency, Accuracy, Processes, Expectations. A set of values I hoped teachers would hang on their walls or drop into a presentation slide so students would have something concrete to hold onto when they reached for an AI tool.

The original framing was student-facing. Students, when using Generative AI... STICK TO TAPE. It was direct, practical, and honestly, a little compliance-flavored. Not wrong, just early. A lot has changed since then. I wrote a book. I spent two years in classrooms and conference halls across Asia testing these ideas with real teachers and real students. And somewhere during those years, TAPE grew up.

“Transparency”

The original graphic asked students to be transparent about when and how they used AI. That still matters. But I missed something important: teachers are users too.

When educators use AI in their planning, their feedback, and their communication, and say nothing about it, they’re modeling secrecy. Students are watching how adults navigate these tools, and if the adults are quiet about it, the implicit message is that AI use is something to hide or manage rather than discuss openly.

Transparency now runs in both directions. It’s a professional practice, not just a student disclosure policy. When you model your own AI use openly, you normalize thoughtful use and invite students into the same kind of reflection. You become a co-learner rather than an enforcer.

“Accuracy” to “Agency”

This is the change I want to spend the most time on because it wasn’t obvious to me at first. Accuracy was always present in the information literacy chapter of my book. Can students verify what AI produces? Is it factually sound? Those questions are still real and still worth asking.

But accuracy is narrow. It asks about the output. It doesn’t ask about the learner.

After two years of developing my writing around AI-enhanced processes, I kept returning to a deeper question that accuracy alone couldn’t hold: who is doing the thinking? That question runs through everything I’ve built. A well-designed learning experience leveraging Think, Generate, Edit, has an answer to it at every stage. The breadcrumbs of student thinking that they leave behind are evidence of it.

The word I needed wasn’t accuracy. It was agency because it asks whether the student is still in control. It holds information literacy inside it, because you can’t exercise real agency over a claim you haven’t verified. But it goes further, into ownership, judgment, decision-making, and what it actually means to learn something rather than generate something.

We’re entering a phase where AI agents don’t just suggest ideas from a chat window but act on our behalf, completing tasks, making decisions, and taking initiative in ways that used to require a human in the loop. That changes what students need from us, because the question is no longer just whether they used AI, but whether they stayed in the work at all. Agency isn’t just a classroom value for learning (like it long has been). It’s the skill of staying the author of your own thinking when the tools around you are increasingly capable of thinking for you.

In another article, I suggested that we think of AI as a guest collaborator. A synthetic guest in our classrooms who can do remarkable things. But the student is the host. The host sets the agenda. The host makes the calls. Agency is what makes that true. You’d be surprised at how often students think they have to comply with AI rather than work with AI.

A great book that explores how AI is less of a tool due to its agency is Nexus by Yuval Noah Harari.

“Process”

Of all the letters in TAPE, P was always the one I felt most strongly about and the one I felt the original graphic undersold.

Process is not a checklist. It’s not a workflow someone hands you. A genuine AI-enhanced process names the kinds of thinking students are expected to do, structures the moments when AI can support that thinking, and captures evidence that real thinking happened along the way. Over time, those structures fade as students internalize the moves and begin to work more independently. That’s the goal. Internalization. Automaticity. Students who don’t need the scaffold anymore because they’ve built the capacity.

Without an intentional process, there are a lot that can go wrong, like AI doing the thinking for kids, not with them. That’s school slop, and it’s what I’ve spent the last few years helping teachers design against.

The documentation “breadcrumbs” concept lives here. Lightweight traces of thinking, a quick voice memo, an annotated draft, a before-and-after comparison, that make the process visible without turning documentation into a second assignment. If there are no breadcrumbs, there’s no way to know whether the student was thinking or whether the AI was doing it for them. And more importantly, there’s no way for the student to know either.

“Expectations”

Expectations were always the structural piece of the earlier graphic, and that hasn’t changed. But the language around them has tightened. Expectations are not rules handed down after a policy meeting, and I felt that the earlier version of the poster might have had that ethos to a certain degree. Expectations in this case are shared agreements, built through clear language, that name the thinking moves expected at each stage of the work. They clarify when AI can support those moves and when it must step back. They specify what evidence will make student thinking visible.

My friend and colleague Nick Soentgerath pushed this idea further in a way that stuck with me: when students and teachers evaluate together what skills and knowledge a task is actually trying to develop, they can make informed decisions together about where AI supports that goal and where it gets in the way. That co-creation has a huge impact on learning, motivation, and the students’ mindsets. Students are far more likely to honor expectations they had a hand in shaping than rules that arrived from above. Shared agreements, built with students rather than handed to them, are the difference between a policy and a practice.

Clear expectations protect agency by design, not by policing after the fact. If a teacher has to chase down evidence of student thinking after an assignment is submitted, the expectations weren’t clear enough upfront. The design failed before the student did.

So here’s where I landed.

The new graphic is called Shared AI Values. I decided to step away from the acronym being the front and center. There was sort of an explosion of AI framework acronyms to the point that it became sort of too much.

In this case, I wanted to think of it as community agreements. The “we” matters because it puts teachers and students on the same side of the same question: how do we use these tools in ways that keep humans, and human thinking, at the center?

TAPE started as a spark in a room in Hong Kong. Years later, it’s a series of values I’ve tested in classrooms across Asia, built into a book, and kept refining every time a teacher asked me a question I couldn’t fully answer yet.

Honestly, the values haven’t changed, but my understanding of them has, and as this technology continues to evolve, so too can the application of our long-held values.

Download the new graphic below. Share it. Adapt it for your context. And if you use it, I’d love to know what conversations it starts.

Shared Ai Values Tape V2 (a3) (1)

7.14MB ∙ PDF file

Download

Download

Monday-Ready Ways to Use This Poster

Student discussion prompt:

“Look at the poster. Pick one value you think is hardest for you during assignments. What would it look like in your work today? Give one concrete example: ‘If I’m protecting Agency, I will ___ instead of ___.’”

Teacher modeling move:

“I’m going to show you what transparent AI use looks like on a real task.”

I’ll project a tiny example:

• My prompt: “Give me 5 possible headlines for this paragraph.”

• What I used: headline #3 (with edits)

• What stayed human: the claim + the final choice

  Then: “Notice what I’m doing: I’m not hiding AI. I’m also not letting it make the decisions that matter.”

Breadcrumb example:

Value Tag Breadcrumb (students add one sentence at two points)

• Midway checkpoint: “Right now I’m using Process by ___.”

• End checkpoint: “The biggest decision I made (Agency) was ___. I can point to where it shows in my work: ___.”

AI Disclosures

One of the most repeatable practices you can bring into your classroom tomorrow is the use of AI disclosure statements. You’ve probably seen them at the bottom of articles. Mine tend to run long (occupational hazard). But a student’s disclosure can be three sentences. The word count isn’t what I want to emphasize; what matters is that it’s honest, and that writing it requires them to actually think about what they did.A disclosure is a breadcrumb. It makes the process visible. And a well-written one touches all four values.

• Transparency is the obvious one. The disclosure gets interesting when students have to answer: where did AI show up?

• For Agency, what did you keep, cut, push back on? Why? If a student can’t answer that, it’s useful data for both of you.

• Process shows up when students describe their thinking moves. “I used ChatGPT to help write this” tells you nothing. “I outlined my argument first, then rewrote the conclusion myself” tells you everything.

• And Expectations close the loop. Did you follow the agreements your class built together? A disclosure is a natural place for students to name that, without you having to police it.

When students write disclosures regularly, they start thinking about their process while it’s happening. That’s the practice we’re after.

Thanks for reading! If you liked this material, please share it with a friend who you think might benefit from it.

Share

Listen to this article

Here is an audio recording of me reading the article below. Listen to it on your drive home, on your phone like a podcast, or follow along while you read.

TL;DR

II’m back after a break for Chinese New Year!

Process-based teaching is meant to protect thinking. With AI in the classroom, it can accidentally protect the appearance of thinking instead. But before any process can hold, the essential conditions for learning have to be in place. This article looks at both through a self-determination lens: the load-bearing walls a classroom needs, and the six predictable ways a process still breaks down even when they’re standing.

Intro

If you’ve seen me speak, you’ve heard me say that AI doesn’t create new problems. It reveals old ones.

Take one that’s been hiding in plain sight: we have long valued the static artifact of a finished essay over the messy, cognitive act of actually constructing an argument. When a student can generate something that looks finished in seconds, old habits get exposed. When the essential conditions for real learning aren’t in place, students learn quickly that what matters is to look finished, not to think deeply, or even to know what that looks like. And when a dysfunctional system meets a tool that can produce polish without effort, it shows its hand.

Process-based teaching is meant to protect thinking. In the AI era, it can accidentally protect the appearance of thinking. And here’s the uncomfortable part: most teachers think of themselves as process people. They believe in the work. They value thinking over polish. The problem isn’t bad intentions or misplaced values. It’s the behavioral implementation gap between believing in process and actually designing for it.

That gap is what separates an AI-enhanced educator from a slop-enabling one. Not philosophy. Design.

This article is a list of the six predictable ways process-based learning breaks down with AI, plus the early look-fors that tell you it’s happening.

Before we start, great process design requires the essentials to be in place first. I didn’t invent those essentials. Deci and Ryan’s “Self-Determination Theory” (SDT) names three of them: autonomy, competence, and relatedness. That third one is worth pausing on, because relatedness is the term that tends to trip people up. In SDT, it refers to the felt sense of being connected to the people around you, of mattering to others, and having others matter to you. But in a classroom context, relatedness fails in two distinct ways that are easy to confuse. A student can feel genuinely cared for by their teacher and still have no personal stake in the work. And a task can connect to the real world and still land in a room where a student doesn’t feel seen or cared for by their teacher.

So I’ve split relatedness into two walls, drawing on Zaretta Hammond’s work in Culturally Responsive Teaching and the Brain: belonging, the felt sense that a student is known in the room, not just present in it, and relevance, my own interpretation of her ideas around cultural connections to learning: the sense that this particular work has something to do with a student’s own life, community, and experience.

So: Belonging. Competence. Relevance. Autonomy. These are the load-bearing walls of a classroom, as I see it. A student who doesn’t feel seen will feel invisible in the learning. A student who finds the task too difficult will use AI as relief. A student with no personal stake in the work will use AI as the fastest way out of what looks like a checklist. Process design without these walls just creates more elaborate hoops to jump through.

So the conditions come first. But here’s the harder truth: even when they’re strong, a process can still fail. Small design mistakes compound fast when “finished work” is one prompt away. That’s what this article is about. Here’s how processes break.

Six Common Pitfalls of AI-Enhanced Processes

1) Vague language leads to vague behavior

Every breakdown in this pitfall starts with unclear communication. When teachers say “AI allowed” without defining what that means, and when schools frame AI as “just a tool” without examining what that metaphor actually teaches, students are left to interpret expectations that were never made explicit. That’s not an academic honesty problem, though many teachers see it as one. For many students, it’s actually a communication problem.

“AI allowed” without a definition leaves students guessing. Some use AI for idea support. Others outsource the whole task. Mismatched voices, “gotcha” moments, and disputes are the predictable result of thinking expectations that were never stated clearly in the first place.

The same problem lives in the metaphors we use in the classroom. Tool language is genuinely useful early on. It makes AI feel manageable, puts responsibility back on the human, and gives schools a practical framework for early questions: What is this for? Who should use it? When does it help? But tools are neutral. Generative AI is not. It pushes back, persuades, and carries patterns from its training data in ways a hammer never could.

Framing AI as a guest collaborator picks up where tool language stops, and it carries a different set of values with it. A guest has a role, operates within norms, and is welcome but not in charge. But more than that, a guest changes the nature of the work. Learning isn’t transmitted from AI to student. It’s constructed through the interaction: the push and pull of a draft that gets questioned, an argument that gets challenged, an idea that gets stress-tested and comes back stronger. The guest doesn’t write your essay. The guest reacts to your draft, pushes back on your argument, and asks what you actually meant. That’s a co-learning relationship that we like and want to continue to instill in young people.

There’s a relevance cost to vague language, too. When task language is generic, it signals that any student could have received this assignment. It doesn’t invite students to bring their own experience, community, or perspective into the thinking, and it doesn’t honor or celebrate the families, values, and identities they carry into the room. A student who doesn’t see themselves in the work has no particular reason to do the cognitive heavy lifting when AI can produce something plausible without any of it.

Spot the pitfall early:

• You might see “AI allowed” in the task brief, and nothing specific beyond that. Fix it with one concrete allowed/not-allowed example and a short why. Then ask yourself one more question: does the task brief invite students to bring their own experience, community, or identity into the thinking? Generic language and vague AI permissions tend to arrive together, and they send the same message: this work wasn't designed with you in mind. Students need to see what your expectations look like in action, and they need to see themselves in the task.

• You might hear students say “I just used it to clean it up,” “you never said we couldn’t,” “just tell me what to do,” or “does it have to be in my own words?” Fix it with precision: name the step in your process where AI enters, and name who owns the decisions. “In this step, use AI to draft two alternative structures. In the next step, the decisions are yours.” And model it first: “Here’s how I used AI this week and what I noticed.” That positions you as a co-learner and shows students what honest reflection on AI use actually sounds like.

• You might notice students describing AI's role in completely different ways in the middle of a task that you thought had clear expectations. That's not a dishonesty problem. It's a norming problem: the class never built shared language around what AI's role actually is in this process. Fix it by making that conversation public. Post a one-sentence class agreement on the wall: “In this task, AI is a guest at the drafting stage. The decisions stay with the writer.”

2) Documentation serves policing rather than learning

A process designed to protect academic integrity isn’t automatically a bad thing. But when documentation serves the teacher’s need to prove honesty rather than the student’s need to see their own growth, the purpose gets inverted. Breadcrumbs that could show a learner how far their thinking has traveled become evidence in a case. A folio that could make effort visible and meaningful becomes a paper trail. And students who already weren’t sure they belonged in the room now have confirmation that they’re being watched rather than supported or coached. They’ll invest in not getting caught rather than in the learning.

The better question isn’t “how do we catch them?” It’s “what have we designed that makes outsourcing feel logical?” If the assignment is product-only and thinking never has to show up during the work, detection won’t fix the design problem. A task that hides thinking produces students who hide AI use. That’s a design flaw, not a character flaw. The stronger move is a process so intentional that AI use is built in, expected at a specific moment, and tied to a specific cognitive move. When the teacher defines where AI enters and what the student has to do with it, there’s nothing to hide. There’s also no ambiguity about where the effort needs to be exerted. A rigorous system of thought (I love that phrase) doesn’t leave room for shortcuts because it already accounts for them.

Spot the pitfall early:

• You might see process steps treated as proof rather than support, and AI use that is hidden or defensive rather than disclosed. Fix it by making transparency the default: share the prompts you used to build a task, the output you rejected, and the decisions you made. Ask students to follow your example.

• You might hear students say “I didn’t know we had to document that,” “the AI just helped me format it,” or go quiet when you ask them to walk you through their process. Fix it by building documentation into the process itself: “Your note should include one place where AI pushed your thinking and one place where you pushed back on AI.”

• You might notice more energy going into catching AI use than into redesigning the tasks that produce slop in the first place. Fix it by co-creating norms with students and naming the purpose: “We document AI use so you can see your own thinking grow, not so I can audit it.” That shifts your role from enforcer to mentor.

3) The process becomes a ritual without meaning

A process with named steps isn’t automatically a thinking process. If students can’t explain why a step exists or what cognitive move it requires, they’ll complete it the same way they’d fill out a form: accurately and meaninglessly to efficiently complete the task. A student can brainstorm with AI, select an idea, and move on without making a single deliberate cognitive move. The brainstorming happened, but the thinking didn’t.

Remember the walls of the classroom in the opening? Well, in the case that students follow a process without effort, it’s likely a competence issue. A student who can follow steps but can’t name their thinking doesn’t feel capable; they feel procedurally compliant. And a student who can’t see what their effort is actually producing has no reason to invest more of it.

One example I have experienced as a student is “reflection”. It was something many of my teachers had me perform over my academic career and while it was a powerful learning strategy, I wouldn’t say I ever understood why it was important or how to do it well. Today, students could easily say: bah, I’ll just get ChatGPT to whip something up and skip this nuisance of a reflection, it doesn’t offer me new insights.

Spot the pitfall early:

• You might see students complete every step but struggle to explain what thinking they actually did, or why the step mattered. Fix it by adding a “name the thinking” prompt at transitions: “Before you move to the next step, write what you just decided and why.”

• You might hear students say “I just did what the step said” or “I wasn’t sure what I was supposed to be thinking about” when you ask them to walk you through their process. Fix it by giving students language they can actually use: “I accepted this because...” or “I pushed back here because...” A shared anchor chart of thinking moves with verbs and brief definitions externalizes the invisible. When students can name what they’re doing cognitively, they’re more likely to do it intentionally. Without that language, thinking stays implicit and the step stays mechanical.

• You might notice transition points passing without anyone surfacing what just happened cognitively. Fix it by talking with students about why certain thinking moves matter: “Why do we reflect? How does it help us hold onto what we just experienced?” When students understand the science behind the step, the step stops feeling like a hoop.

4) Documentation as post-mortem autopsies

This pitfall is related to the second one but distinct from it. Pitfall 2 is about why documentation exists: when it’s designed to catch rather than coach, it becomes policing. This pitfall is about documentation happens: even well-intentioned documentation fails if it arrives too late. A teacher can genuinely want to support student growth and still fall into this trap by collecting everything after the work is done.

This is where autonomy matters, and where formative assessment becomes the practical alternative. Documentation that lives after the work is done belongs to the teacher’s evaluation, not the student’s learning. But documentation designed as connected tasks during the work, each one building on the last, creates something different: a visible record of thinking developing over time. That’s where feedback does its most powerful work. Not as a judgment on a finished product, but as a response to thinking in motion. A teacher who can see a note about a decision from day one, an annotation from day three, and a concept map from day five can give feedback that actually moves the learner forward. Breadcrumbs are invitations for formative response.

Spot the pitfall early:

• You might see process artifacts that only exist at submission, with no thinking trail built during the work itself. Fix it by requiring tiny breadcrumbs in the moment: “Annotate one place in your draft where you disagreed with AI and explain what you did instead.”

• You might hear students say “I already finished it” when you ask for a decision note, or notice that the only documentation in the room is happening at submission. Fix it by building checkpoints during work time: “At the halfway point, drop a decision note: what have you changed since you started and why?” Breadcrumbs built during the work are invitations for formative response.

• You might notice grades that still signal product matters and process is optional, with no weight given to how thinking developed over time. Fix it by including a short “defend your choices” moment: “In your pair share, explain one decision your breadcrumbs show that your final product doesn’t.” Consider delaying the grade until the process is complete and feedback has been acted upon.

5) The process is too rigid or too loose

Under pressure for certainty, processes start to resemble algorithms: fixed sequences meant to produce predictable results. Every student does the same steps in the same order at the same pace. But the opposite failure is just as common. A process with no real structure or access to the teacher leaves students guessing what the teacher wants, and when students are confused, AI becomes the fastest way to resolve the uncertainty. Both failures produce School Slop (this term refers to student work that is AI-generated rather than AI-enhanced; output that looks complete but reflects no meaningful cognitive effort.) One produces slop through compliance, while the other through perceived abandonment.

The balance isn’t a middle point between structure and freedom. It’s a design that places cognitive load on the thinking rather than on figuring out what’s expected. A well-designed process gives students clarity about where they are, what they’re doing next, and where they can get support. It doesn’t predetermine where they’ll end up. It scaffolds the journey while leaving room for the thinking to go somewhere real.

This is an autonomy and competence problem. A rigid process communicates that student judgment doesn’t matter. A loose one communicates that the teacher isn’t invested enough to be present. Both erode the walls of the effective classroom. What students need are frequent check-ins, strategic feedback at the right moments, direct teaching, and a clear understanding of how AI supports the class's goals without replacing the cognitive work that builds capability. It’s so much more than content.

There’s also a longer-term goal on the horizon worth designing for. The goal was never compliance with a process. It was students internalizing the thinking moves behind it, adapting them to new contexts, and eventually designing their own. A student who has genuinely absorbed a process doesn’t need the scaffold anymore because the scaffold has become instinct. That’s also where wise AI use lives: not in following rules about what AI can and can’t do, but in a student who knows their own thinking well enough to know when AI is enhancing it and it’s replacing it.

Spot the pitfall early:

• You might see every student moving through the same steps at the same pace, regardless of where their thinking actually is, with no room for a different interpretation to land. Fix it by designing for clarity about expectations, not predictability of outcomes. Students should know what’s expected at each step, but the process should leave genuine room for their thinking to go somewhere unexpected. Turn steps into decision points: “If your draft is coherent, move to revision. If it’s messy, use AI to propose two structures, then justify which one you chose and why.”

• You might hear students say “I don’t know what you want” or “can I just use AI to figure out the next step?” Fix it by being present during the work and building in frequent check-ins: “At each checkpoint, I’ll spend two minutes with each group. Come ready to tell me where you’re stuck.” Strategic feedback at the right moment matters more than a perfectly sequenced process. Make it explicit how AI supports the class’s goals: students should be able to answer what AI is helping them do and what thinking stays theirs.

• You might notice students who can follow the process but can’t explain why any step matters, or students who turn to AI the moment the structure gets ambiguous. Fix it by planning for fading explicitly: “By week four, students design their own process for this task type using the thinking moves we’ve practiced together.” We want students to internalize the thinking moves well enough that the scaffold becomes a natural part of how they operate.

6) How we assess learning

This is where every other pitfall converges. Undefined AI use produces polished work with no thinking behind it. Policing replaces the coaching that would have caught it early and kept thinking with the students. Silent steps mean students completed the process without understanding it. Post-mortem documentation gets retrofitted to match the product. Rigid and loose processes push students through without checking whether understanding is being built.

And now, at the end, if the only thing being evaluated is the final product, every one of those design failures gets rewarded and reinforced with a grade. A rubric that can't tell the difference between a thinking student and a polished AI draft isn't evaluating learning. It's evaluating presentation. This is a relevance problem: when the grade doesn't value thinking, the assessment tells students what actually mattered, and it wasn't their ideas.

Thinking = effort.

When thinking doesn’t show up in the grade, students get a clear signal: the cognitive work doesn’t count. And a student who has figured out that thinking doesn’t count will find the fastest path to a product that looks like it does.

The fix is a holistic evaluation: criteria that only a thinking student can meet. Justified decisions. Evidence of revision. Original connections that held up in conversation. A student who can explain their choices in a two-minute conference is demonstrating something a polished AI draft never could.

Spot the pitfall early:

• You might see AI-first drafts becoming final drafts because revision is optional, and process evidence that gets collected but carries no weight in the grade. Fix it by grading process and product, even lightly, so thinking has perceived value: “Twenty percent of this grade is your decision trail: where you changed direction and why.” Make revision count explicitly. A revised draft with a rationale earns more than a polished first attempt.

• You might hear students say "I don't remember why I changed that" or "the AI suggested it and it seemed better" when you ask them to explain a decision in their own work. Fix it by building a process journal from the start so thinking is captured as it develops, not reconstructed after the fact. Example assignment prompt: "Alongside your final product, submit a journal that shows where your thinking started, where it changed, and where you landed." Example conference prompt: "Open your journal. Walk me through one moment where you changed your mind."

• You might notice a rubric that rewards polish and completeness more than decisions and reasoning, where a polished AI draft scores the same as a well-thought-out one. Fix it by adding criteria only thinking can meet: “Your rubric includes one criterion your final product alone can’t satisfy: explain one original connection you made that AI didn’t give you.”

Monday-Ready Design Checklist for Process-Based Learning with AI

Each pitfall above comes with a concrete fix. If the look-fors resonated, here’s a free downloadable checklist PDF that pulls all six into one diagnostic you can use before your next learning task with AI in the room. Print it, post it, share with a colleague, bring it to a planning meeting with your team, or keep it on your desk.

Process Design Checklist

115KB ∙ PDF file

Download

Download

Five Readings Worth Your Time

Self-Determination Theory, by Edward Deci and Richard Ryan, is the research backbone behind the load-bearing walls I describe in this article. What I appreciate about their work is how cleanly it names the psychological conditions for human motivation and growth: autonomy, competence, and relatedness. It’s a simple framework that holds up in classrooms every day.

“Tomorrow is Today: The classroom I’m afraid of”, by Alex Kotran, is one of the clearest pieces I’ve read on AI in education. He warns that AI doesn’t automatically create equity for all; in fact, it can entrench inequity if we keep defaulting to a “pedagogy of poverty” and skip serious teacher training.

10-25, by David Yeager, connects nicely to the Carol Dweck mindset work I’ve mentioned in other posts. Yeager is a contemporary and collaborator of Dweck’s, and he makes a great point about what young people are scanning for constantly: belonging and respect.

The Paradox of Choice: Why More Is Less, by Barry Schwartz. This book connects directly to the autonomy wall and pitfall 5. Schwartz makes a counterintuitive point that shows up in classrooms every day: more options do not automatically create more freedom. Too many choices can overload students, increase second-guessing, and push them toward avoidance or shortcuts.

Culturally Responsive Teaching and the Brain, by Zaretta Hammond, reframes relevance as a neurological condition, not a motivational one. Hammond’s argument is that the brain won’t take on the cognitive risk of hard thinking unless it first feels safe and seen. If students don’t sense that a task connects to something real in their lives, the brain treats it as low-stakes noise and looks for the path of least resistance. That path, right now, is AI.

AI Disclosure

This article took 9-10 hours to write. There was a lot of effort put into it in the form of brainstorming, researching, revising, word-smithing, editing, and creating visuals.

The ideas are mine, mostly drawn from my book and previous articles as well as the literature that I listed in the reading section above.

I used Claude Opus and Sonnet 4.6 and Gemini 3 Pro with Deep Research to test my thinking against the broader literature, then worked section by section with AI as a collaborator. I found Claude Sonnet 6.2 did the best job with understanding the intentions and giving me feedback on how to adjust and make it digestible for readers. Then I used ChatGPT 5.2 Thinking to check that it still sounded like me. ChatGPT was sort of my top-level editor that I ran everything by to make sure it all made sense and the thread running through the article was clear.

Thank you for reading my article! If you liked it, please share it with someone who you think would enjoy its content.

Share

See Me Speak

This academic year, I have a few more public speaking engagements lined up where I will be speaking about the above ideas and more. I hope to see you in person and talk more about AI-enhanced processes!

• 21 CL, Hong Kong - Breakout sessions

• AIFE, Yokohama - MC and breakout sessions

• AI in Action, Beijing - Keynote and breakout sessions

Inviting AI into your classroom and simply saying, “Use it responsibly,” is like inviting a wacky stranger to your dinner party and saying, “Just don’t make it weird.”

Nobody knows what that means. You turn your back for five minutes to grab the wine, and you come back to find the stranger reorganizing your spice rack by scent and trying to sous-vide the family cat. Everyone is staring at you, and you’re left scratching your head, trying to reverse-engineer the chaos from the mess.

That’s the problem with vague instructions: they invite interpretation. As Katie Novak points out in the foreword to the forthcoming second edition of my book, AI-Enhanced Processes, being a host isn’t just about opening the front door; it’s an intentional design. In a classroom, we’re hosting a digital guest in a room full of humans. If we don’t facilitate the introduction, the humans don’t know how to engage, and the guest, with all its enthusiasm and desire to please, takes over the party learning experiences.

That’s basically the current state of AI school policies. Often, they're labels that sound decisive but rely on conceptual language. Many teachers are still unsure how to put that language into practice. Don't get me wrong, schools need a policy that establishes a shared language for how stakeholders will engage with AI in classrooms. However, action plans, practice, professional development, and coaching are the necessary next steps. Without the practical implementation plan, we likely will encounter selective interpretation: a teacher thinks “use AI only for brainstorming” when indicating AI is allowed. A student hears “drafting with AI is fine.” A school leader hears “future-ready learning.”

We need a better way to host AI as a guest in our classrooms, and that means practicing our shared agreements, not just writing them.

That’s where an AI-enhanced process (AIEP) comes in as one meaningful strategy to operationalize good practices. AIEPs are short, repeatable sequences of thinking moves that make expectations concrete. They keep the effort with the learner and give AI a clear role without letting it take over. Instead of “AI allowed,” we get shared language for action: here are the thinking moves, here’s how AI can collaborate on each one, and here’s what stays human.

A strong AI-enhanced process includes these facets by design:

• Multiple steps that break a complex task into smaller moves, so learning happens through a sequence.

• A variety of action-oriented thinking moves (generate, analyze, evaluate, reflect) that build toward a larger outcome.

• Visible thinking through drafts, notes, prototypes, and reflections so the work can be coached while it’s still happening. Process journals, folios, photos, Padlets, to name a few, are great ways to document student learning. It’s most powerful when it’s for the students first, and adults second. I will share more about this below, but I also have a two-article series you might enjoy. They touch on big ideas like efficiency and effort. Links below.

• Explicit roles so each step clarifies what students do and how AI can collaborate without taking over like our wacky guest at the dinner party.

• Embedded AI literacy happens through classroom dialogue and intentionally designed learning experiences, where AI shows up in ways that actually enhance learning. Students learn by doing. They build meaningful habits with AI when lessons use it to strengthen thinking, not replace it. Reflection helps students notice how the process actually flowed for them: where AI supported their growth, where it got in the way, and what they want to adjust next time. Over time, that cycle of practice and reflection becomes a lifelong strategy that students can carry into adulthood.

In summary, AI-enhanced processes are strategies we intentionally design in conjunction with meaningful classroom dialogue about how we use AI with our students, which can lead to powerful learning experiences and lifelong strategies.

Ok. Now that we have shared language for what an AIEP is, we can actually host AI in our classrooms. In the next sections, we’ll look at three powerful moves to build your own processes that take the above ideas into account and operationalize them.

• Powerful Move 1: Name The Thinking

• Powerful Move 2: Set Expectations around Student-AI Collaboration

• Powerful Move 3: Make Learning Visible

Powerful Move 1: Name The Thinking

Naming the thinking in a process is an idea borrowed from Project Zero’s work around making thinking visible. And in that sense, this is not really a new idea; it’s the notion that we consider the category of thinking, and then name the verb we want students to practice. For example, the category might be to think critically, and the verb might be to evaluate. Project Zero suggests that if you care about thinking, you have to design for it. You have to name it, value it, and make it observable, because what a classroom repeatedly reinforces and rewards is what students learn to do. And with AI, if we don’t name the thinking, the easiest path of the least resistance wins: creating a polished product automatically. It’s tempting for students to skip the messy process that requires effort and to focus on task completion. If we do name it, we can build the habits of mind we actually want to instill in our students.

Another important thing to note when naming the thinking is that each thinking move is not an “activity” to keep kids busy, nor are they compliance checklists. They’re steps that make thinking visible enough to practice another powerful move: metacognition, in which we can talk about the thinking, coach for improvement, encourage students to notice their own growth, and plan their own independent next steps.

In addition to Harvard Project Zero’s Thinking Routines, there are other processes that follow very similar patterns, like Stanford D.School’s Design Thinking cycle, writing workshops, lab reports, Socratic seminars, critique protocols, coaching cycles, research, and a hundred other classroom processes that work because they support the thinking that leads to a solid product.

Below is a preview of my book with samples of words that we can draw upon. It’s not an exhaustive list, but it’s a good place to start, helping us think about what we mean by “naming the thinking moves”. Notice the bolded words under the “Example Thinking Verbs” column. Those are things that AIs can currently do quite well on their own or in collaboration with our children.

When planning, ask yourself:

• What types of thinking are important in this task?

• If I were to name what students should do as verbs, which ones come to mind?

• How can I build a shared vocabulary with my learners about what these words mean?

• When in the year is it helpful to introduce, practice, and review this language?

In the next section, we will take the named thinking moves and set expectations for when and how AI collaborates with students on each move.

Powerful Move 2: Set Expectations around Student-AI Collaboration

I like to frame AI as not a tool, not a calculator, and certainly not a threat, but a guest collaborator. That framing changes the way we introduce it to students. We’re not handing them a device and saying, “Go use this.” We’re inviting AI into the learning space, and we should be deliberate about the roles it plays.

For example, if the thinking move is brainstorming, you might say: “AI will collaborate by asking Socratic questions that push your creativity. It will not generate ideas or suggestions for you.” Now students know what they’re responsible for, what AI is responsible for, and what’s off-limits.

You can take that one step further by narrowing which AI students use. In class, that might sound like: “I built a bot for you to use that’s aware of our learning objectives. Please use this bot only. It helps me be a better teacher, it allows me to coach you with more precision, and it makes your AI use transparent so we’re both clear on when and how you leveraged AI to support your effortful thinking.”

From there, the planning move is straightforward: decide how AI supports each thinking move based on your knowledge of the learning goal and your students. Caveat: this kind of decision making does require some AI literacy from us as teachers, enough familiarity to predict what the tool will do well and where it might take over the thinking.

If you’re new to it, AI-use scales can make expectations easier to name and share. You can use an existing one or build your own.

One option is the Expectations Scale that Holly Clark and I created. It helps you label what AI can and cannot do within each step of a process. For example: “During empathizing, you may use AI to give feedback on the depth of your interview questions,” or “During outlining, you may use AI to help you structure your plan.”

Another option is the AI Assessment Scale by Perkins et al. It’s widely used and has influenced a lot of school-based scales. The concept overlaps with what you saw above, and to be transparent, our work was influenced by theirs. Their team updates the scale regularly and grounds it in ongoing research, which makes it well worth reviewing as you design expectations for students. Check out more here: https://aiassessmentscale.com/

Ok, so now you’ve seen two moves: name the thinking, then decide how AI will collaborate with students, or not, on each step of the process.

In the next section, we will build on those two ideas with a third powerful move. In the next section, I will argue that a process only works if we can actually see it. If the only thing we ever see is the final product, we’re back to guessing how the students arrived at that work and praying that AI didn’t do it all for them.

The third powerful move is making the learning visible through documentation. Let’s take a look.

Powerful Move 3: Make Learning Visible

I realize that up until this point it has sounded a lot like I’m placing process over product. Let me be clear: the final product still matters. For most kids, it’s the most motivating part, the destination, the why of it all. But when a compelling product sits on top of visible thinking, it becomes something richer: a story of learning they can actually explain. Look at what I made, and here’s how I got there. I struggled. I persevered. I pivoted. I grew.

The problem is that AI can generate polished work instantly. So a polished product alone is weaker evidence of thinking than it used to be. That’s why making the thought process visible is powerful. Meaningful learning requires both process AND product. Products give purpose. Processes shape the thinking that makes the product worth creating. Drafts, prototypes, revisions, annotations, and reflections are mini-products that capture thinking along the way.

But there’s a trap that teachers should be aware of: documentation can slide into policing. When visibility becomes prove you didn’t cheat, students experience it as compliance, so I use a better frame: breadcrumbs. Breadcrumbs are lightweight traces of thinking captured during learning: a quick reflection, a sketch, a revision note, an outline, a screenshot of a chat plus a short takeaway. Not everything, just enough to make learning visible in context and help students notice their own growth over time.

There are countless ways to document student thinking. Here are two visibility moves that work across subjects, then a list of more possible approaches. If you would like to see more inspiring ideas, make sure to check out my book.

Option 1 - Process journals:

A process journal gives students a single place to capture “breadcrumbs” of their thinking as it evolves, and it gives teachers something coachable while the work is still happening. It also scales cleanly: a simple template in Google Docs, Word, or Pages can be duplicated for every student through an LMS, then adjusted up or down with prompts, feedback boxes, links to vetted tools, and clear expectations. The result is one adaptable tool that holds instruction, links to purpose-built bots, reflection, and evidence of learning together, instead of trying to reverse-engineer what happened from a polished final draft. In fact, final drafts can sit in the process journal! It’s such a powerful move to ask students to submit their process journals to their teachers that include the final draft because it shows that we value the thinking and the work leading up to the final piece.

If you like the idea of using a process journal, I have templates and more ideas that you can get for free from these two articles.

Option 2 - Think Collaboratively on Paper

Vertical learning, the Peter Liljedahl move you’re seeing pictured below, is a simple design choice with a huge payoff: get students standing, working in groups, and thinking out loud on visible, shared spaces.

In the age of AI, this technique is powerful because it shifts the class from “who can produce the cleanest final answer” and back to how ideas form, change, collide, and improve while simultaneously building in movement. My goodness, do kids sit a lot during the day! Vertical work makes an active process unavoidable: students get up, and start talking about learning away from distractions like laptops.

If you notice, the laptops are closed in this classroom! Students go from discussing, collaborating, debating, and then perhaps opening their computers to share their findings with an AI bot from their teacher. Such a powerful and intentional use of technology at key times.

More ways to document student learning

I only shared two approaches in this article, but there are plenty of ways to document learning. Below is a list of additional options students can use to capture their thinking as it’s happening. The goal is simple: make learning visible enough that you can coach it, students can learn from it, and AI stays in the role of collaborator. You’ll probably mix and match a few of these rather than rely on just one.

Possible approaches include:

• Blank template/drafting space (work-in-progress thinking).

• Chat thread (shows thinking changes, not just answers)

• Pivot point notes (moments thinking shifted and why)

• Metacognition question trails (curiosity evolving into sharper questions)

• Idea graveyard (discarded ideas kept as evidence of exploration)

• Connection notes (new idea + prior knowledge)

• Annotated snapshots (image + one sentence of meaning)

• 60-second voice memos (messiest parts + next moves)

• Evidence selection (best evidence set + short rationale in slides)

• Physical notebook entries (kids love these)

• Conference/exhibition script (student-led explanation supported by evidence)

How to Design Your Own Process

One of my favorite ways to build an AI-enhanced process is to use intentionally low-tech tools: sticky notes and a pen.

To be clear, the stickies are for planning. They’re a coach’s sketchpad. They’re perfect for teams because you can move them around fast, test different sequences, and talk through how students will flow from thinking move to thinking move without getting stuck in a Google Doc spiral.

When I’m coaching a teacher, I’ll make two rows. Row one is the thinking moves. Row two is how AI collaborates, or doesn’t, on each move. That’s it. You can see a quick example in the photo above from a high school writing task. I made it up for this post, but you get the idea.

Once you’ve got a sequence you like, share it with students in whatever format fits your classroom. Take a picture and project it. Turn it into a slide. Write it on the board. Print it and put it on tables. The format doesn’t matter. The point is simple instructions that keep the focus on the learning and make expectations easy to follow.

Teachers, I gotta say, me included, we talk too much. We frontload instructions like we’re reading the terms and conditions of a credit card. We overexplain the pitfalls, the rules, the exceptions, the “don’t do this,” the “also don’t do that,” and by the time students actually start working, their brains are already tired. That’s cognitive load in the worst place: on the directions instead of the thinking!

A simple process fixes that because it keeps instructions short. A process can enable you to provide just-in-time teaching. You can circulate, coach the move students are actually in, and drop a mini-lesson, a metacognitive guide, or feedback right when it’s useful. Direct teaching is not bad; in fact, it is one of the highest impact approaches to teaching, according to Hattie. But when it’s the predominant teaching move, class can become disengaging and focused on content, compliance, and control.

Here’s the sticky note move I use when coaching teachers to plan their own AI-enhanced process:

• Step 1: Write thinking moves across the top row (verbs only).

• Step 2: Under each one, write the AI expectation in plain language: “AI can help by… / AI cannot… / Student must…”

• Step 3: Share it in class. Read it once. Ask for clarifications. Then start the work. Coach during the process and avoid over-explaining, as tempting as it is!

Pro tip: take it one step further and add a third row for breadcrumbs. Under each thinking move, write what students will capture to document their learning and make the thinking visible, for example: a brainstorm snapshot, a decision point, a revision note, a reflection, or a screenshot of a chat. Whatever it might be, it helps you and your students to align around what thinking will look like.

Monday-Ready Resources and Powerful Moves

There were a lot of ideas in today’s article. What I want to leave with you with are big picture, powerful moves, and further tools you can start using in class immediately.

1) Good reads

The Power of Making Thinking Visible, by Ron Ritchhart and Mark Church. If you want a practical backbone for naming thinking and making it observable, this is a strong place to start.

Mindset, by Carol Dweck, has had a powerful influence on my work with AIEPs. Dweck explores the concept of a growth mindset versus a fixed mindset and how a fixed mindset can limit our growth and potential. This mindset shapes many areas of our lives. What Dweck does so well is help us recognize the language of a growth mindset. With AI in our classrooms, we need to pay close attention to the language of a fixed mindset. I’m seeing many ways a fixed mindset could lead to students leveraging AI to complete tasks rather than seeing that productive struggle is an important and necessary part of the learning experience.

2) My free process builder

You can use my tool to generate an AI-enhanced process and a detailed prompt you can feed into a chatbot so it understands your context, your students, and your documentation approach. It helps you to not only get specific about the learning experience you want to design, but it also helps AI to align with how it will support because you can feed the prompt from this tool straight into School AI, Flint, Magic School, etc., to follow your process. https://processes.lovable.app/

3) The three takeaways from the article.

These are the three things I really want you to walk away with. Immediate and simple steps to support AI use in your classroom with clarity.

1. Name the thinking.

2. Set Expectations around student-AI collaboration.

3. Document student thinking and make it visible.

Conclusion

If you want a starting point, don’t overhaul everything. Redesign one task as a process, and watch what happens when expectations stop being implied and start being practiced. Start small on purpose, and start where you already have momentum. Think about the repeated thinking moves you see all year: brainstorming, outlining, arguing from evidence, checking for bias, revising for clarity, or reflecting on choices. If there’s a strategy you use frequently, that’s your best candidate. Operationalize it. Name the moves, make them teachable, and run the same process again and again until students stop asking, “What do you want?” and start asking, “Which move are we on?” or even better yet, show independence and fluency with their use of the process.

Write a short AI-enhanced process for that task, and keep it simple:

• Name the thinking (verbs only).

• Set expectations for how AI collaborates or doesn’t.

• Add breadcrumbs so the learning is visible while it’s still alive.

And that’s it. You’re not building a system. You’re building a routine that students can actually run without you reading the terms and conditions out loud for 30 minutes.

“AI allowed” is a label that invites interpretation, and interpretation is exactly how you end up with a spice rack sorted by scent and a cat in a sous vide bag.

A process turns shared values into shared behaviors. It protects the parts of learning that can’t be outsourced: judgment, reasoning, taste, decision-making, and reflection. And it gives students something better than “What do you want?” It gives them a map that reinforces your shared values. After frequent practice, the questions shift to the ones you actually want to hear: “Which move are we on?” Or even better, no question at all. They just start, and they run it. The point being that processes need to be regularly reinforced to get to automaticity and internalization. With AI, kids have few internalized processes due to the fact that it’s a relatively new technology, so your efforts to practice meaningful use of AI are essential.

While I believe in the power of process, to be clear, the point is never the process itself. The point is the thinking and the effort that the process makes visible.

Start small on purpose. Run the same process again and again. Tweak one step at a time. Find ways to demonstrate understanding. Keep the focus where it belongs: on student thinking, not on policing, not on overexplaining the steps, and definitely not on cleaning up the mess after your digital guest takes over the party.

AI Disclosure

Supporting artwork in this post was created with the support of ChatGPT and Gemini. All images have AI disclosures to indicate my process with transparency. Other graphics from my book were created by me (e.g. table of words).

The audio accompaniment of this article with me reading was my voice, and then run through Adobe Voice Enhancer to clean it up and make it sound professional.

This article was co-written by ChatGPT 5.2 and Claude 4.6. My process followed several cycles of Think, Generate, Edit. This article is in many ways a summary of the core ideas in my book. I took a PDF of it (and compressed the hell out of it), uploaded it to ChatGPT, and had a conversation with it about my work. We created an outline that followed it and the typical structure of my Substack posts.

My typical flow is to start an outline on a Sunday, think during my workweek, especially while walking to work and talking to my partner (walk-and-talks are the best). Then on the weekend, I refine the work for Monday morning to be delivered to your inboxes.

What I found is that ChatGPT and other LLMs are great at helping me write, as I can get a summary of my existing work/ideas and start laying it out. I’ll wake up early, have coffee, and read where I am in my writing, revise it through a conversation with my synthetic collaborator, and then go about my day. I find new ideas come to me, and I will edit the work later to help it flow, or sometimes even to visualize it better.

I would estimate that, with AIs as my collaborators, I saved a significant amount of time, but I still spent around 10-11 hours completing it.

Finally, I believe we should normalize transparent disclosure of AI use in education. When adults model openness, students are more likely to talk honestly about how they’re using AI rather than hide it. If you are not disclosing your use of AI, you might be unintentionally modeling secrecy.

Thank you for reading.

-Alex

P.S., It’s Break Time

I’m going to take a break for two weeks from publishing a weekly article for the Chinese New Year, to travel a bit, and do some thinking. I’ll be back again at the beginning of March for more great content, including a podcast with Dr. Sabba Quidwai, the pitfalls of process-based education, and more.

If you liked this article and want to keep up with my writing, consider subscribing or, better yet, sharing this article with someone who might enjoy it. I really appreciate your support!

Share AI Enhanced Processes

We’re three years into the AI experiment, and I’ve noticed something during that time: the way we talk about AI keeps shifting. The metaphors change from month to month, and they shape the conversation more than we realize. A lot of authors and speakers pick one analogy and argue that it’s the best way to understand what AI is and what it means for education. In this meta-article, I want to do something different. I want to look at the range of metaphors we use, unpack what each one highlights, and consider their limitations.

This article isn’t trying to crown a single “right” metaphor. It’s a case for building a metaphor toolkit we can use deliberately. Metaphors are helpful because they make complex ideas feel manageable. But they also push our thinking in certain directions. They can narrow what we notice, shape what we assume, and set limits on the questions we think to ask. This HBR article provides data that very well may back those ideas up. What it shows is that our metaphors matter in how we work with AI.

So today I’m going to look at five categories of AI metaphors. For each one, I’ll explore how it helps our thinking, how it limits our thinking, and when it’s useful.

Category 1: Calming Metaphors

This is perhaps the first form of metaphor that I heard being used in 2022 when ChatGPT exploded onto the scene. This section is dedicated to metaphors meant to address skepticism or reluctance.

Common phrases you’ll hear in this category of thinking are:

• “We’ve been here before, it’s like when everyone freaked out over the calculator.”

• “ChatGPT is just the next Google Search or Wikipedia, remember when those were new?”

1) How this category helps our thinking.

One metaphor I hear a lot is the idea that we’ve been here before, and the comparison people reach for is the calculator. It’s one of the metaphors that we can probably laugh about the most because it’s considered such a common technology at this point. Check out this newspaper article with protests against calculators in lower elementary school.

Unpacking AI as a “calculator” metaphor, it’s often used to lower the temperature. When ChatGPT started writing essays on demand, it rattled assessment overnight and raised big questions. What does it mean to “know” something? What evidence counts? What is the role of feedback? The calculator comparison gives educators a familiar storyline: a disruptive tool shows up, schools adapt, and eventually it becomes normal.

And calculators aren’t the only example. Schools have pivoted before in plenty of other ways. We pivoted hard during COVID when learning moved online. We pivoted again when 1:1 programs put a device in every student’s hands and laptops became part of daily classroom life. Those shifts were messy, uneven, and sometimes frustrating, but we adjusted.

2) How this category limits our thinking.

The calculator comparison breaks down in two important ways.

First, calculators are narrow tools with predictable outputs. They don’t generate ideas, rewrite arguments, imitate voice, or steer a student toward a particular conclusion. Calculators don’t do relationships. Even Joaquin Phoenix wouldn’t date a TI-83.

Second, calming metaphors can turn into minimizing metaphors. If we treat AI like “just another calculator moment,” we risk missing what’s actually new here: a system that participates in language, produces plausible explanations, and can shape student thinking while sounding confident. That creates different risks and demands different literacies. Verification, source judgment, and attention to persuasion become central, not optional.

This is why, as Dr. Sabba Quidwai has argued (paraphrasing), AI isn’t “like a calculator.” A calculator executes. Generative AI responds. Sometimes it’s brilliant, sometimes it’s wrong, and it can be persuasive either way. If we stop at “we’ve been here before,” we may calm people into staying put when they actually need to learn new skills and redesign curriculum, assessment, and norms with eyes open.

3) Suggested use of this type of metaphor.

I would use this trope sparingly when a community is anxious and new to AI, as a way to reduce panic and open dialogue. Avoid it when the goal is serious planning, because it can oversimplify what makes generative AI uniquely powerful and uniquely risky.

Calming metaphors are helpful because they reconnect us to a belief: we can handle this. They’re especially useful at the beginning of the AI journey, when people are panicking and need to remember, “we’ve got this.” The takeaway is that we need to move through our first emotional reaction so we can return to a sense of efficacy and begin a real dialogue with students about how AI is changing the world they’re growing up in. But calming metaphors become a problem when they’re the only way we talk about AI. The calculator metaphor, for example, isn’t really a claim that the technologies are alike; it’s a reminder of a familiar moment: disruption, panic, adjustment. Useful, yes, as long as it moves us past knee-jerk reactions and into the work of learning, adapting, and having meaningful classroom dialogue about what AI is for, how it’s being used, and what still needs to remain human. As you’ll see in the sections that follow, other metaphors can help us operationalize AI in practice instead of getting stuck in the emotional first response.

Category 2: AI as A Tool

Common phrases you’ll hear in this category of thinking are:

• “It’s a tool.”

• “AI is like a knife. It’s dangerous when misused, but it’s powerful when used appropriately.”

• “A hammer won’t build the house for you; it’s the carpenter’s skill that matters.”

• “AI is like a car, and you need a license to know the rules of the road.”

1) How this category helps our thinking.

Tool language is useful because it makes AI feel manageable. Teachers understand tools. Tools are chosen for a purpose. Tools belong in a workflow. This metaphor helps schools move from panic to practice by asking practical questions: What is this for? Who should use it? When is it helpful? What is the cost? It also moves responsibility back onto humans. A hammer does not build a house by itself, and it does not get credit or blame. The person using it does. That mindset can be grounding when people are tempted to treat AI output as truth or treat the model as the decision-maker.

Tool metaphors also make space for safety. We already teach students that powerful tools come with rules. You do not hand a first grader a saw in your makerspace. You teach basic use, supervision, and responsible handling. AI fits that same pattern. If we treat it like a tool, then we naturally ask about purpose and function, training, and age-appropriate use.

2) How this category limits our thinking.

The tool metaphor can also oversimplify what generative AI actually is. Tools are neutral and depend upon the user. Knives don’t talk back or persuade you to pursue flawed ideas, while Generative AI does, and in an often biased way, depending on its corpus’s training data.

Tool language can also push people toward a vending machine mindset. Put in a prompt, get out a product. That is where shallow work lives and where Prompt-Copy-Paste becomes the default. It can also hide the fact that AI is not a neutral tool. It carries patterns from its training data, and that can show up as bias and distortions. In short, when we call it “just a tool,” we can accidentally train students to treat output as something you either accept or reject, rather than something you interrogate, negotiate, or iterate upon. More on this in the 5th category.

3) Suggested use of this type of metaphor.

Use this metaphor when your goal is to build basic readiness. It is especially helpful for setting guardrails, introducing responsible use, and clarifying that humans own the goals and the final decisions of AI output. Pair it with clear limits so it does not turn into AI is just another thing to “do school” and get a grade, so we can move on.

Category 3: AI as A Threat

Common phrases you’ll hear in this category of thinking are:

• “The weaponization of AI could turn society into Foucult’s Panopticon.”

• “We’re letting the genie out of the bottle.”

• “AI is Pandora’s box.”

• “We’re unleashing the Terminator or even Frankenstein’s Monster: our creation that’s going to turn against us!”

• “We are in an AI arms race!”

1) How this category helps our thinking.

Threat metaphors are useful because they surface many risks that we should consider, like surveillance, manipulation, data misuse, deepfakes, disinformation, and inequality. They also remind us that systems do not need to be evil to cause harm. Sometimes harm comes from convenience, lazy policy, or unintentional adoption.

In education, threat language helps us protect what matters most to us: students are learning and growing into good humans. It pushes us to ask hard questions about student privacy, consent, and what data is being collected. Threat metaphors raise the alarm on dependency and cognitive offloading, where students lose the habit of wrestling with ideas because an answer is always available. It also calls attention to power. Who controls the tools? Who benefits? Who gets left behind? How’s it impacting our environment?

2) How this category limits our thinking.

When we frame AI in the language of threats, schools often respond with bans, blanket rules, and fear-based messaging. The result is that students hide their AI use, and it becomes harder to teach them how to use it well. Some students start to equate any AI use with cheating, even when it’s being used in a productive, transparent way.

Furthermore, there’s a difference between a student using AI to get feedback on writing and a school deploying surveillance-style analytics on every click. If we only speak in the language of threats, we miss the chance to build AI literacy, which is the best protection we have.

3) Suggested use of this metaphor.

Use this metaphor when you need to raise standards and slow down adoption. Avoid using it as the main classroom narrative for students, especially younger ones, because fear is not a foundation for good judgment.

Category 4: Productivity

Common phrases you’ll hear in this category of thinking are:

• “It’s an amplifier.”

• “AI is a spark for my ideas and thinking.”

• “It’s magic… and can you time as a teacher.”

1) How this category helps our thinking.

Productivity metaphors are popular because they are often true. It can speed up drafting, brainstorming, differentiation, translation, and formatting. For teachers drowning in work, this category of language resonates. It also helps people see that AI can amplify good practice. A strong teacher with clear goals can use AI to generate options faster and then apply professional judgment.

This category also helps with momentum. Schools need quick wins to get started. Productivity is the quickest win. Used well, it can free teachers to spend more time on relationships, feedback, and the human work of teaching.

2) How this category limits our thinking.

These metaphors can reinforce a culture where efficiency is the goal. In classrooms, that plays out when students use AI to skip the messy middle: the struggle, the uncertainty, the revision. The output looks impressive, but the thinking is superficial.

At the school level, the same mindset can shift us toward quantity over quality. More lessons, emails, and rubrics. More content. If we’re not careful, AI becomes a machine for producing more school slop (low-quality content nobody wanted write, and nobody wants to read, including the person who initiated it).

This framing also downplays risk. When the primary goal is saving time, people verify less, reflect less, and rely on shortcuts. And we shouldn’t assume the “time saved” automatically turns into more time building relationships with learners. In many schools, it simply becomes the capacity to do more administrative work, because the workload expands to match whatever efficiency we create. After all, we don’t have more contact time during our prep blocks.

3) Suggested use of this metaphor.

Use this metaphor when your goal is teacher workload relief or rapid iteration, especially behind the scenes. Be cautious using it as the dominant student-facing metaphor, because it can habituate students to optimize for speed over depth. Remember, adult use of AI is different than student use of AI, and how you speak about and model AI is important.

Category 5: AI as Teammate

This is one of the most common metaphor categories I hear in schools right now. It shows up whenever people are trying to move the conversation past “AI is just a tool” and toward something more like shared work, shared thinking, and shared responsibility. All of the examples below have something in common: they use humanized relationship and role language. They also encourage us to think less about output and more about how AI fits into the learning process.

Common phrases you’ll hear in this category of thinking are:

• “AI is an intern.”

• “It’s a co-pilot.”

• “AI is your collaborator.”

• “It’s a thought partner.”

• “It’s a coach or personal trainer; those people don’t do the work for you, but they help you build muscles. AI can do that for your mind like a tutor.”

1) How this cateogory helps our thinking.

Jeremy Utley from Stanford University talks about this category of AI metaphor a lot. His video and follow-up blog post are linked in the “Where to Learn More” section below. One message of his that sticks with me is simple: do not use AI, work with it, treat it like a teammate. The best part starts around the six-minute mark.

The teammate metaphor encourages us to think in terms of delegation. In any good team, we do not let one member do all the work while everyone else watches. We assign roles, we make expectations explicit, we make sure each person is an active contributor, etc. That matters even more with AI, because we are effectively placing expert-level intelligence next to a developing learner. AI can outpace a student’s thinking simply because it can generate faster, wider, and with more confidence than most kids can.

So the question is not “Can AI do this task?” Of course it can. The better question is “Who is doing what, and what thinking still belongs to the student?”

At the same time, teammate language does not make AI harmless. I actually appreciate the tool metaphors above because they encourage respect: knives, hammers, and fire are useful in the right hands and dangerous in the wrong ones. But we are not dealing with a tool waiting to be picked up, nor are we dealing with a predictable calculator that follows a linear sequence. We are dealing with emergent intelligence that can associate, remix, and improvise, and in the not-too-distant future, likely feel, reason, and have a level of sentience.

That is why the real skill is purposeful collaboration as a form of AI literacy and fluency: an ongoing conversation with pushback, follow-up questions, and feedback. That mindset helps us avoid unhealthy processes like Prompt-Copy-Paste (it sounds like a dystopian Thinking Routine, doesn’t it?).

“I don’t use AI. I work with AI.”

2) How this category limits our thinking.

Collaboration language can blur accountability, just like we see in group work: students sometimes hide behind more active partners. AI can become a similar cover unless expectations are explicit within the learning process. A recent Harvard Business Review study of over 1,200 managers puts some numbers behind that notion.

When AI was framed as an employee rather than a tool, personal accountability dropped 9% and workers were 44% more likely to escalate decisions to a manager rather than stand behind their own judgment; in other words, fewer errors were caught and people blamed AI for the mistakes. The framing metaphor changed the relationship and behavior. One line from the research stood out to me: “In an agentic system, how work is divided between humans and AI is a design choice.” That is the argument I keep making on this website and in my book. As teachers and students with agency, we must deliberately decide what belongs to the student and what belongs to the AI. Steps like naming the thinking, sharing expectations, and deciding what counts as evidence of learning (i.e., what I often call breadcrumbs) all have a positive impact on student learning when working with AI collaboratively.

Furthermore, teammate language also encourages over-anthropomorphizing. When AI sounds human, it is easy to treat its outputs as intentional or authoritative, which I call AI omniscience bias. When we believe in AI’s ultimate intelligence, we lose our healthy skepticism, making it harder for students to practice verification and careful judgment. AI can also dominate the direction of a task unless the process is designed to protect student agency and voice.

So in short, if using teammate language, it’s essential to remind students of their ownership of the content and to practice judgment, skepticism, and critical thinking.

3) Suggested use of this type of metaphor.

Use this metaphor when your goal is to promote students using AI in specific and intentional ways in the learning experience. It is especially useful for teaching students that AI can contribute to brainstorming, feedback, revision, and practice while students remain responsible for the thinking and decisions.

To use it well, pair it with explicit norms:

• Define roles for AI and roles for humans before students begin work.

• Create purpose-built bots that help students use AI while following the intended process.

• Name the thinking moves that must remain human and where AI is allowed to collaborate.

• Require evidence of student reasoning and decision-making, not just a finished product (i.e., document the process).

A Monday-Ready Conclusion

Choose your metaphors like you choose your instructional moves, based on the moment and the outcome you want. Use calming metaphors when your community is anxious, and you need to lower the temperature enough to begin a real conversation, but do not let “we’ve been here before” become an excuse to avoid updating curriculum and assessment. Use tool metaphors when you need clarity and guardrails, especially when setting expectations and shared responsibility. Use threat metaphors when you need to slow down adoption and take risks seriously, but do not make fear the main student-facing story, or you will push use underground. Use productivity metaphors when the goal is workload relief and rapid iteration, but be cautious. Efficiency can become the point, and AI can turn into a machine for producing “school slop.” Use teammate metaphors when you want purposeful collaboration, where the process is clear, roles are defined, and students are expected to be active learners who take responsibility for the AI output. Going back to the HBR article, that means checking the content and standing by what they put their names on.

If I had to add one more metaphor to this toolkit, it would sit inside the teammate category: AI as a guest in the room. When we invite a human guest into class, we do due diligence. We think about fit and purpose. We plan the visit so it supports the curriculum. We prep the guest, and we prep students. Then, when the guest arrives, we introduce them. We explain why they are there, and we name how we are going to work with them. We set simple norms that keep the focus on learning.

That is the posture I think we need with AI. Even if it is “the same tool” each time, AI is a chameleon. It can take on different personas and play different roles depending on how we frame the interaction. One day, it is a brainstorming partner. Another day, it is a feedback coach. Another day, it is a debate opponent, a summarizer, or a tutor. Because of that, teachers have to be deliberate about the introduction each time we invite AI into the room. What role is it playing right now? What is the learning goal? What does collaboration look like in this moment, and what does it not look like? If we skip that framing, we leave too much open for interpretation, and students end up making their own calls about when and how AI should be used. That is where misunderstandings happen, and that is where “dishonesty” can get murky fast.

To make this practical, we have to get more specific than “AI allowed” or “AI not allowed.” Those labels create gray areas and invite confusion. A better approach is to name the thinking we want to see, then indicate when and how AI, as a guest in the room, is allowed to collaborate on particular thinking moves across an assignment. Break the larger task into the moves that actually produce learning. Be explicit about which moves must remain human and which moves can be supported through collaboration. Keep the focus where it belongs.

These metaphors could be a signal of where we are on the trajectory of understanding AI in education. Everyone is free to use the metaphor that fits their current experience, and early on, calming or tool metaphors may be exactly what a community needs. But if you have been using AI intentionally with students and you feel further along in your thinking, this is an opportunity to push the current collaborator narrative one step further. You can expand it by leaning on an existing practice we already understand well: how we vet guests. In this case, we are moving from an organic guest to a synthetic one. The point is to treat it as a guest that deserves respect and careful planning to protect students and their intellectual growth. Thank you for reading.

Where to Learn More

• Jeremy Utley’s Substack & Interview on YouTube. This is the author’s Substack and the interview that I refer to in the Teammate section of this post. I really appreciate his thinking and work around creativity, as well as his rationale behind why metaphors like collaborator positively impact our interactions with AI.

• Mindset, by Carol Dweck. As my friend Emily would say, this is an evergreen book that I believe is especially relevant to this post. Our mindset and the ways we think about learning are reflected in the metaphors we use. If you have not already read this powerful book, I highly recommend it.

• “The Adolescence of Technology”, by Dario Amodei. This essay came out this week and has taken the internet by storm. Amodei is the CEO of Anthropic, the company behind Claude. This lengthy essay is rather dense, but it is worth your time in both its content and its use of metaphors to understand AI. In the essay, there are several great metaphors for AI. Amodei compares the AI era that we are in currently to an adolescent growing up. He also talks about a future with powerful AI as being a country full of geniuses in a data center, along with many more.

• “Students Are Skipping the Hardest Part of Growing Up”, by Clay Shirky. This is an opinion post from the New York Times and is focused on the social impact of AI on young people. The article is full of metaphors for AI, but my favorite that I hadn’t heard before is that AI is a cigarette, implying long-term costs from our AI use to our social relationships.

• Conferences. This academic year, I have a few more public speaking engagements lined up where I will be speaking about the above ideas and more (see more on my Conferences page of the site). I hope to see you in person and talk more about AI-enhanced processes!

  • 21 CL, Hong Kong - Breakout sessions

  • AIFE, Yokohama - MC and breakout sessions

  • AI in Action, Beijing - Keynote and breakout sessions

• Subscribe. If you haven’t already subscribed to this newsletter, hit the button below! Or if you know someone who might benefit from it, please share it and help me grow my audience.

Subscribe now

AI Disclosure

All origami images in this post were created with the support of ChatGPT and Google Gemini to illustrate the ideas in each section.

It’s also worth sharing that the reading of me at the very top of this article was me recording a reading for 30 minutes of the post and then enhancing it with Adobe’s Voice Enhancer, an AI model that is quite good at making everyone sound like they’re a talk show host. What I found is that the higher quality the recording, the better the results.

I co-created this post with ChatGPT and Gemini. It took me 8-10 hours to write because of the thinking, revision, and back-and-forth collaboration. The LLMs’ roles were to push back on my ideas, help refine my writing, flag redundancy and verbosity, and help me maintain my tone and style. I ran a report after finalizing the entire article to fact-check the piece to ensure there were no glaring or errant claims. The thinking, final edits, and publishing decisions were mine.

Finally, I want to end by sharing my values. I believe we should normalize transparent disclosure of AI use in education. When adults model openness, students are more likely to discuss their AI use honestly rather than keep it secret. My goal is to engage in dialogue with the students, and that means it’s essential to avoid a judgmental atmosphere. Modeling how to use AI to enhance our thinking is one way to open the lines of communication.

I share my process because I want students to develop independence and wise decision-making with all technology. And it’s fair to say that most teachers use AI without disclosure. So, in many ways, you might actually be modeling secrecy.

In this episode, I have three chats with different international educators who are working with AI and assessment in different contexts. My previous episode on assessment was one of my more popular, so I thought it was time to come back and see where we were at in terms of thinking that might be developing or getting more refined. It’s been a year since we recorded the last episode. Wow, time flies!

Let’s take a look at the details of what you can expect and the folks joining me in order of appearance in the show.

Emily J. Thomas is an educator, educational consultant, and entrepreneur who supports international schools in strengthening curricular development, coherence, and a clear vision for teaching and learning. She has spent over a decade in IB international schools as an MYP/DP English language and literature teacher and, most recently, served as an MYP Coordinator; she’s also an IB Educator Network workshop leader and a DP Literature examiner, and works as a literacy strategist with Erin Kent Consulting (EKC). Alongside her work in schools, Emily founded Playground Pedagogy (“playful minds, serious learning”) and leads yoga-focused work through Teaching Matters Yoga and Drift Yoga in Bangkok, and she writes the weekly Substack Elsewhere, Examined.

In this conversation, Emily reframes assessment as an opportunity to extend learning; a way to “tune in” to what learners have actually acquired, not a checkbox to end a unit. She unpacks why formative vs. summative terminology can create anxiety and mixed signals for students and argues for schoolwide clarity, including shared definitions, consistent language, and policies that treat formative evidence as meaningful rather than “worthless.” Turning to AI, Emily’s message is “process first”: the best response is doing the fundamentals well with simple, standardized task sheets and clear expectations (including what AI use is appropriate) that teachers and students see consistently across classes. She closes with empathy for educators navigating this moment and a call for leaders to “steer the ship” with clarity so teachers can feel calm and supported.

Timothy Cook is an educator and the founder of Connected Classroom, exploring how AI shapes student cognition and learning. He currently teaches third grade at the American Community School in Amman and writes Psychology Today’s “Algorithmic Mind” column, where he examines the intersection of education, AI, and human cognition, especially the risks of dependency and what schools can do to protect critical thinking, creativity, and moral development.

In this conversation, Tim argues that writing still matters more than ever because it’s fundamentally a process of thinking: the focus, word choice, revision, and self-argument that helps students clarify what they actually believe (and that AI can’t authentically replicate). He introduces the idea of “jagged edges” that include the human, lived, imperfect uniqueness that gets flattened when AI produces the same “academically average” response to predictable prompts. From there, he makes a practical case for “AI-proofing” assessment by redesigning tasks around community, identity, and design: prompts where students must apply content in locally grounded ways (and where AI can still be used as a tool without replacing the thinking).

Nick Soentgerath is a Technology Learning Coach at Yokohama International School (Japan), where he supports teachers and students in designing practical, future-focused learning with a strong emphasis on ethical, responsible, and safe use of AI.

In our conversation, Nick brings a practical, classroom-grounded lens to what assessment can be when it’s less about “gotcha” grading and more about clarity, feedback, and growth. Helping schools move from measuring learning to actually improving it. He also presents at international conferences and works with educators on assessment practices that are more authentic, equitable, and aligned with the skills students need beyond school. In the episode, Nick and I discuss the upcoming conference at his school. Find out more here: www.AIFE.community.

Provide Feedback about This Episode

Graphic designers, musicians, photographers, and copywriters have found their crafts threatened by low-quality “AI slop”: AI-generated content that lacks the human touch but gets forgiven because it amuses us, or because of the hype surrounding the technology, that we accept what would normally be unacceptable.

Ok, if you can get the image of Shrimp Jesus out of your head, let’s go back to creativity, efficiency, and process.

An efficiency-first mindset becomes risky when applied to creativity, because creativity is not just “making something”, and it is especially not whipping something together and calling it a day. The type of creativity we want students to engage in is actually a challenging mental task. I’ll speak to this more in a moment with a graphic below, but creativity is highly personal and expressive; it’s originality and collaboration; it is analysis and technical skills.

Some phrases you might hear that would indicate this sort of efficiency-first mindset is at play are below. Also in the conclusion, I have these look-fors with alternatives that we want to hear instead. Some preview samples include:

“I just need to get this done.”

“What’s the minimum to get an A?”

“Do we have to show our process?”

“ChatGPT, write this essay for me, here’s the prompt.”

“I’m stuck, can you just tell me the answer?”

“Why do we have to do multiple drafts?”

“Let’s have the students all do the same type of project.”

If the priority is simply finishing a task, AI starts to stand in for the creator instead of supporting the human work of expression. The path of least resistance will always win, and if the goal is task completion rather than meaningful thinking, then in the age of AI, we end up with “school slop.” In schools, that often means assignments no one feels invested in: students generate the work with AI, teachers respond with AI, and a grade gets passed back and forth. It is fast, it looks finished, and it drains the human presence that makes creative learning worth doing. It’s compliance, not creativity, and while this existed before AI, AI is certainly shining a light on this problematic mindset.

In this dystopian premonition of AI taking over creative thinking and relationships, we get into a cycle in which teachers assign work that doesn’t merit effort, so students stop believing effort matters. Students then stop trusting the teacher’s expertise because the feedback feels automated, too. That human connection, the part that was so important, gets diminished due to innovation for innovation’s sake. School slop becomes normalized thanks to AI and a curriculum that might not have kept pace with the rapid changes in our world.

In the upcoming second edition of my book, AI-Enhanced Processes, I dive deeply into AI superpowers. Here is a preview of three concepts from the new book that frame how we might protect creativity and avoid school slop.

Classroom dialogue ideas

Can AI be creative, or is that something only people can do? How can we still express ourselves and invite AI into the space without sliding into “school slop”?

Concept 1: The “Steak Dinner” Theory

I am incredibly fortunate that the second edition of my book features a foreword by Katie Novak, a leader in Universal Design for Learning (UDL). In her foreword, Katie shares a brilliant analogy about product versus process. She compares education to hosting a dinner party. It doesn’t matter how good the steak or the wine is (the product) if your guests are vegetarian or sober. The goal isn’t the food itself. The goal is the gathering and each person being nourished (the process and the outcomes).

This analogy frames what I am trying to do in the second edition. We are moving away from looking at the shiny AI tool or the product as output (the steak) to designing the experience (the dinner party). When we apply an efficiency mindset to a dinner party, it fails us. We don’t want to efficiently consume calories; we want a meaningful experience. Learning, in this case, is less about gobbling down information and more about savoring, discussing, connecting, and noticing what students prepared.

School slop happens when we confuse the steak for the party. We get obsessed with output and forget that the point was the experience and nourishment.

Concept 2: Creativity with AI is a “Sandwich”

This post really is leaning on food metaphors. We’ve got “slop,” dinner parties, and now a sandwich. I know, I know. But this one is worth keeping.

If good teaching is nourishing, it should be experiential and cognitively stimulating. The question is: how do we structure lessons so students want to spend effort on thinking, not on decoding confusing instructions, or bypassing learning to complete tasks efficiently and using AI to create slop?

One option I use in the book is Think, Generate, Edit (TGE). The idea with this process is that you should name the thinking moves that matter in your context and build a process around them. Better yet, help students design their own processes once they’re ready. But if you want a simple starting point, try this sandwich.

Think (Human)

We start with a human thinking independently. This step is AI-free to protect productive struggle and asks students to think deeply about something. Thinking in this case could be a drawing, a prompt, a storyboard, a photo, a song, an outline, a draft, etc. For example, in an art class, this might look like: think by creating a sketch of your idea.

Generate (AI + Human)

We use AI through a back-and-forth conversation. We are not doing this to save time. You aren’t using AI as a vending machine; you are using it as a collaborative partner. In an art, this could look like: upload your sketch, generate several interpretations in different styles, refine prompts in detail based on what you notice, then choose the best ones you would like to use in your next step.

Edit (Human)

We end with a human touch by editing, refining, and curating. We never just take AI output and submit it. In an art, this could look like: load the image into an iPad and add your own edits through illustration and effects. This is where school slop gets defeated, because the student is making taste-based decisions and doing real craft.

To illustrate TGE, I share a story in the book about remixing a song I wrote back in 1997.

As shown above, I took a collaborative approach. In this case, a collaborative approach looked like both AI and I contributing to a new and unique song. I uploaded a song that I made as a kid (think), then had a conversation with AI and generated several covers that were more refined with additional instruments (generate). Finally, I edited the song using GarageBand, added effects, and removed sounds (edit). If I had used generative AI to quickly make a song without engaging in any effortful thinking, I might have written a half-hearted prompt in Suno (AI platform that makes music) and called it a day. I’ve seen kids do the same thing, basically telling it to “make a song” without much thought, effort, or process to support their thinking and effort. Without the process, the amusing output from Suno would probably have been meaningless slop that had stopped at entertainment.

Making a song with TGE was so inspiring that it led me to create several other songs, including ones in other languages, featuring instruments I can’t play (keyboard and guitar player here), and vocals I could never produce.

Classroom dialogue ideas

Can AI support your creativity? How should or shouldn’t we use AI in a given learning task? What process could we construct to deliberately use AI in a specific way in our work?

Concept 3: Defining the Creative Learner

Finally, we have to look at who the learner is in this equation. In Chapter 1 of my book, I share a graphic called “Traits of Creative Learners Who Use AI.”

The graphic visualizes three rings:

1. The Center: Our identities, who we are.

2. The Second Ring: Our values.

3. The Outer Ring: Behaviors, and how we use AI when driven by our identities and values.

The point of all of this is that we have a definition of how we use AI (our behavior) when it is driven by our values and identities. The graphic also defines creativity in four domains: synthesizing, analyzing, collaborating, and upskilling. I thought that it was worthwhile to look at how creative thinking is actually quite broad and that it’s not just thinking about making something original.

A second point that is implicit in this graphic and chapter is that creativity isn’t an innate gift (I’m channeling my inner Carol Dweck). Creativity is a rigorous way of thinking that takes effort and discipline, two things the efficiency mindset encourages us to avoid. School slop is what happens when we treat creativity like it’s something we’re born with instead of something that requires real work and effort that is actually fun and rewarding.

And if we aren’t careful, lazy AI-use can bypass the traits in the graphic:

• We bypass synthesizing by prompting and sharing raw output from AI models. School slop loves unedited output.

• We bypass analyzing by accepting outputs without critique. School slop hates critical thinking.

• We bypass collaborating by focusing on the chatbot instead of teamwork and teacher mentorship. School slop is lonely work.

• We bypass upskilling by letting AI do the heavy lifting and skipping the practice that builds craft. School slop is allergic to practice and effort.

So far, we’ve looked at what students need to avoid, but I want to take a moment to mention some of the important things that students do when being creative. First, students need the right physiological state. They have to be calm and alert. If they are stressed, distracted, or tired, then they are not going to access those creative parts of their brains. They need to be physically taken care of first before they can exert meaningful effort.

Second, they need the right mindset: they have to be willing to spend energy because the purpose feels valuable. All humans are hardwired to want to avoid effort and conserve energy, and thinking deeply spends that precious energy. So when students have voice, choice, personal investment in what they express, a tie to their community and culture, an authentic audience to celebrate their work, or any other clear reason why they are creating the work, they will be willing to exert effort instead of looking to efficiency.

Mindset and our physical state are essential to get right before we look at ways we can use AI in our creative processes.

Classroom dialogue ideas

How do you define creativity? What do you value when it comes to creativity? How might the way we use AI be consistent with those values? Maybe that means not using AI at all when we are creative! How are you feeling today; are you calm, alert, and rested? Are you ready to be creative?

Monday Ready Tools

Tool 1: Checklist

Efficiency has its place (admin tasks for those who have already gained expertise, sparking ideas, cognitive sparring partner). But the process is where the magic happens, especially for cognitively demanding tasks like creativity. Clear processes with clear expectations are also how you combat school slop. And teachers, you can design these processes tomorrow. Start by naming the thinking you want students to do, then decide clearly how AI does (or doesn’t) support those moves. Talk to your students about expectations. If you don’t name the thinking, school slop will start to fill the vacuum.

Here is a simple student-facing checklist you can steal and use on Monday:

Think, Generate, Edit Checklist

• Think (Human): I created an original starting point (sketch, outline, plan, storyboard).

• Generate (Human + AI): I explored multiple iterations and can explain what changed and why.

• Edit (Human): I made meaningful revisions and can point to my decisions.

• Documentation of Learning and Process: I can show evidence of my thinking evolving (drafts, screenshots, notes, version history, reflection) and can use them to have a conversation with others about how I grew.

Tool 2: Creative Mindset Look-fors

In the introduction, you saw me write about efficiency-mindset vs. a creative-mindset. Here is an expanded list of phrases you can share with your class when practicing creative thinking.

• Instead of “I just need to get this done” or “Can we just finish”

  “I want this to sound like me. What is MY goal for quality and craft?”

• Instead of “What’s the minimum to get an A” or “How long does it need to be”

  “What does strong work look like to me, and what are the success criteria?”

• Instead of “Do we have to show our process?”

  “What evidence of thinking would best show my creative effort?”

• Instead of “Can we just use the Canva template?”

  “Can I use a template as a starting point, then make original choices and edits that show taste and intention?”

• Instead of “ChatGPT, write this essay for me.”

  “ChatGPT, help me generate options, challenge my thinking, and improve structure. I will write the final draft.”

• Instead of “I’m stuck, can you just tell me the answer?”

  “I’m stuck, but I know I can handle this challenge. Can you ask me questions to provoke my thinking?”

• Instead of “Why do we have to do multiple drafts?”

  “What is the focus of each draft, and what should improve each round?”

• Teachers: Instead of “Let’s have students all do the same type of project.”

  “Can we keep the criteria consistent, but let students choose the format that best fits their idea?”

• Teachers: Instead of “Can’t beat them, join them. We’re all using AI anyway.”

  “Let’s define what good AI use looks like, model it rather than keeping our use of AI secret as adults, and make expectations clear in each step of the process.”

Where to Learn More

I will be diving much deeper into AI-Enhanced Processes and creative moves at several upcoming conferences in Asia. I keep my speaking engagements up to date on the Conferences page on this site. I’ll be speaking in the 2025-2026 school year at:

• EdTech GZ

• AIFE Yokohama

• 21CL Hong Kong

And finally, watch out for the second edition of my book! After all the great feedback from readers and conference attendees, the book has taken on a life of its own. It is a huge leap forward, and I can’t wait for you to read it, coming soon, so stay tuned.

Summary: From Part 1 to 2

In Part 1 of this series, we explored the risk of our students sliding into “the humans of Wall-E” habits in which they use AI for efficiency. In this follow up, we will look at how to operationalize a K-12 classroom that leverages AI to support meaningful work as a “superpower” while still requiring a final product.

A False Dichotomy

I’ve been mentioning the idea of a polished product quite a bit in this article, so it’d probably be helpful to acknowledge that I am not suggesting that we only focus on processes in our classrooms. It’s sort of an age-old argument and it is really a false dichotomy. We cannot have a process without a product, evidence of learning, or artifacts. And we cannot have a product without a process that leads us to creating something worth sharing.

I find that often with kids, the product becomes the why, but they lack the patience or skills to utilize a process to get to a polished product. That is one impactful way teachers can facilitate learning, the co-construction of meaningful processes to support effort, creation, collaboration, and polished products.

Since generative AI is fabulous at creating products, we cannot emphasize only the product, even inadvertently (for example, “here’s a prompt, go home and write on your own, then share your finished work with me for a grade”). A clear process allows the teacher to set expectations and ensures that the student remains the pilot. It breaks something massive into digestible steps. This is what I mean by processes turning into strategies. They help kids practice to think with AI and when we release them to the world after high school, they will have specific ways to use this technology in their utility belts. Wow, so many superhero references, but you get the point.

Solely focusing on a product, without insight into how students arrived at their thinking, can turn into AI helping kids “do school” by efficiently generating a product to satisfy a grade. However, simply adding “process” to an assignment is not a panacea, and should not replace other forms of good teaching.

Furthermore, if we are not careful, process-based education can easily turn into policing where students are overburdened with endless mini-products to demonstrate learning (read this great article by university professor, Jason Gulya, about this idea). In the age of AI, many are turning to process-based education so each step can be documented to prove a student did not “cheat.” This is an autopsy of work, not evidence of learning. When a process is used primarily for surveillance, it becomes compliance rather than a mechanism for thinking. And thinking is the whole point!

In the upcoming second edition of my book, I argue for a shift from documentation autopsies to breadcrumbs, which are contextually natural ways to trace thinking.

Their purpose is to:

• Help students notice their own cognitive moves and support metacognitive awareness of growth. This is a huge one and the priority because it will actually lead to students being increasingly willing to exert effort by seeing that they are learning and growing and noticing that it’s paying off.

• Support teachers in monitoring student learning as it progresses.

A breadcrumb is a form of mini-product that could be a quick voice note, a chat screenshot, a rough concept map, or a sketch. The goal is for students to capture their thought process in a way that supports their learning during and after the experience and to increase willingness to continue to try hard and think deeply. There are so many ways to create breadcrumbs that show a trail of thought. Below are a few ideas, and definitely check out the second edition of my book to see more real-world examples:

• Screenshots Disagreement (A zero-prep approach): Ask kids to capture a screenshot of their chat that contains something from AI they disagreed with and to share their own stance on why it was wrong.

• Screenshot Before and After: Ask the kids the take a screenshot before they worked with AI and then again after they worked with AI and to explain their changes.

• Visual Timelines: Create a Padlet timeline that tracks the steps students took.

• Integrated Journals: Use a process journal with specific spots to draft ideas and interact with a teacher’s bot.

• Frameworks: Follow an established process (like the Stanford d.school Design Thinking cycle) and log ideas in a slide deck.

• Persistent Chats: Have an ongoing conversation with a bot about the process, return to the same chat later, and eventually share the link with the teacher.

A well-designed process that constrains ways to naturally collect artifacts of learning, or breadcrumbs, is my overall recommendation. I would avoid documentation of student thinking that does not ultimately serve the student in reflecting on their learning, or serve the teacher in meaningfully coaching students to do so. Oversight into the thought process as a form of academic integrity is the byproduct and not the purpose.

To make a powerful process, it should explicitly name the thinking required at each stage (for example, “now we are testing assumptions” or “now we are synthesizing viewpoints”) and introduce AI to support that specific cognitive move. The risk of not naming a thinking move and failing to set expectations through dialogue is that students are left guessing about when and how they can use AI. They might overuse AI, or use it in a way that is inconsistent with your goals and be perceived as “cheating.” Naming the thinking moves clarifies expectations for both you and your learners, preventing misunderstandings.

Another overarching goal of K-12 education is to scaffold and support them until they have the agency to design their own workflows and processes. Teachers can provide structures so students can develop the discernment to manage their own thinking later and that likely will emerge as they have more experience in AI as a meaningful collaborator in high school or university.

I want to close this section by emphasizing that breadcrumbs do not have to be a burden. Not every breadcrumb needs to be evaluated. If we treat these artifacts as natural snapshots of thinking that serve the student first, it releases the pressure to grade every single step. A powerful, time-saving move is to have students reflect on their own breadcrumbs at the end, and then simply assess that reflection rather than the piles of rough work. Furthermore, breadcrumbs of thinking act as a safety net in that teachers can dip into them specifically when they notice gaps in a process or work that doesn’t sound like the student’s voice.

Classroom dialogue ideas

What would be possible ways to make your thinking visible during our process? How might you know if you grew during this experience? What would be a reasonable use of your time to document your thinking breadcrumbs along the way? What things would you like to be mindful of when using AI in each step?

Monday Ready Tools

If this has felt a bit theoretical, here’s a practical move you can use in class on Monday.

Goal: design a simple process that (1) names the thinking, (2) sets clear AI expectations, and (3) leaves small breadcrumbs of learning.

Steps

1. Name the thinking move (process). Alone, with your class, or in groups, list the thinking needed to succeed on the task. Use as many steps as you need.

   Examples: generate ideas, plan, draft, check evidence, revise for clarity, reflect.

2. Set AI Expectations (AI-enhancement). For every thinking move, decide how AI can be used, or where it cannot. This is where “AI literacy” becomes practical: What tools do we have access to, and how do we use them so that they support (not replace) the thinking?

3. Collect breadcrumbs of evidence (formative assessment). Pick a small, natural artifact that helps students notice growth and helps you spot when AI did too much. Examples: a before/after sentence, a screenshot of disagreement with AI, a quick voice note (“what I changed and why”), a draft, a sketch, a brainstorm mindmap.

Option 1 to Make an AI-Enhanced Process

Access material I’ve made for you on AIEnhancedProcesses.com > Downloads has a series of posts with things you can use in your classroom to create a process as a teacher or collaboratively with your students. Below is one tool you can use that focuses on writing processes. The image below shows purple cards that are process word. Underneath them are expectation cards that show to what extent AI can be used in that step of the process.

This tool is but one of many in the downloads section of my site. I’ve also got a Padlet template that does the same thing as physical cards, but it’s all virtual and free, making it easy to distribute. But what if you can’t print my materials or perhaps they’re not suitable for your classroom. I’ve got you.

Option 2 to Make an AI-Enhanced Process without Cards

If you can’t print cards, simply make your own with sticky notes. I’ve done this in a pinch many times when coaching teachers, and it worked.

• Row 1: thinking moves

• Row 2: AI expectations (Allowed / Limited / Not allowed + one sentence why)

• Row 3: breadcrumb choice

The main point is the habit: involve students in the process, and encourage them to adapt it with rationale. That’s how we build self-directed learners who make wise decisions about AI, problem-solving, and learning.

Classroom dialogue ideas

Let’s make a bank of words that show thinking; what should go in it? Given our learning task at hand, what thinking moves should we use to create our shared process? If you’re ready, what thinking moves could you sequence to create your own process? What existing processes are out there that could work in this situation or be adapted to meet our needs (e.g. Design Cycle)?

Classroom Example

In Anita Vanessa Chou’s MYP 5 science class at Yokohama International School, AI bookends a scientific investigation with Foresight and Hindsight as two forms of metacognition. Before the lab, a Flint chatbot helped students clarify the purpose of the learning task, restate expectations connected to Criterion B, break the objective into variables and constraints, surface questions and uncertainties, and plan a safe investigation with intentional procedures.

Below is an excerpt from my book, AI-Enhanced Processes, in which I do two things: I label the steps in which AI was used within a deliberate process with a little robot. Second, I name the “superpower” that AI supported in the learning with a hero. Let’s take a look.

After the pre-task foresight, the students conducted the lab. The point is that the thinking stayed with them. They still had to make decisions, interpret what they were seeing, and deal with the messiness of real science. AI did not do that part for them. It was also highly collaborative between students, which matters because collaboration is one of the easiest things to accidentally erase when AI becomes a default shortcut.

After the lab, the same tool supported metacognition by helping students review their report and feedback, notice patterns in variable control and explanations, hypothesize why those patterns occurred, and choose concrete strategies to improve next time. The AI did not do the thinking for them. It scaffolded the thinking they were meant to practice as scientists, mostly through Socratic questioning that encouraged metacognition before and after the learning task.

It sounds counterintuitive, but using AI probably made the whole experience slower. But hey, that is a very good thing. That lazy and critical-thinking part of our brains I mentioned earlier, in Kahneman’s book, needs time and friction to actually do its job. In this case, AI did not accelerate the work; it increased the fidelity of the thinking.

Classroom dialogue ideas

What did you see happening in this classroom that shows AI was a collaborator with the students and not an assistant completing things for the learners?

Conclusion

If there is one thing I want to leave you with, it is this: efficiency is not the goal of students who are learning to use AI. The habits we practice with kids today will become the habits they practice as adults. So the question isn’t whether students will use AI. The question is what kind of habits we are building? Wall-E habits, or Iron Man habits?

Now what do we do with that? We stop treating AI like a blanket permission or a blanket ban, and start treating it like a design choice inside a learning-first process. In every assignment, we can name where the thinking workout needs to stay with the learner (productive struggle), and where AI can remove barriers that don’t actually create learning (unproductive struggle). We ask students to keep breadcrumbs to serve themselves and their learning first, and second to check whether AI is supporting the thinking or creating cognitive debt; the quick diagnostic shared above can help determine if students engaged in the learning. The breadcrumbs should be small, natural traces that help students notice their own cognitive moves and help teachers coach learning as it progresses. Not to police kids.

We don’t just want polished products, we want AI to be a collaborator that strengthens judgment, confidence, and independence, the stuff that actually transfers when students are on their own.

Classroom dialogue ideas

What superpowers does AI make possible for us when it comes to animals, science, medicine, transportation, knowledge? If you were to imagine an ideal world, how would you like to see AI become a superpower for you as a learner?

Great Reads Related to This Article

These are some of the works I reference, allude to, or was inspired by while writing this article. All of them look at the ideas presented here from a different angle and are recommended reading in the form of articles or books.

• “Problems with ‘Process Over Product’ (Part 1)” by Jason Gulya

• “Why Does Thinking Feel So Hard?” by Carl Hendrick

• “Gotcha Culture” by Colleen Ferguson

AI Disclosure

Part 2 of this article was created through an iterative Think, Generate, Edit process, repeated multiple times. The initial draft was written by me and informed by my work on AI-Enhanced Processes, Second Edition. I then used multiple AI models: ChatGPT, Google Gemini, Claude Sonnet, and ChatGPT Deep Research. They were used to refine, challenge, and extend my thinking. Visual elements were generated with ChatGPT. All writing and final decisions were reviewed and edited by me, a real human, to ensure coherence, intent, and accuracy.

It’s also worth disclosing how long it took me to write this article. I would estimate 8-10 hours were spent writing, thinking, rewriting, generating, etc. I would not say that was a particularly fast way of working, rather AI was my collaborator.

At a Glance

Students often arrive at AI-assisted research with an assumption: that AI is all-knowing, omniscient, and accurate. I call this AI Omniscience Bias, and it’s the trap Generate, Verify, Reflect is designed to address. Students use AI to find or synthesize information, verify what they find against other sources, then reflect on their own assumptions about accuracy. That last step is the one that builds AI literacy; it turns research into a metacognitive habit.

Generate

Overall, what we want to see students doing at this stage is being critical of information and AI models in the moment; this step is not for us to look at other sources, which will be completed in the second step of the process. Here, AI is the source, and students are learning to treat it like one.

1. Use the ROBOT Test to evaluate the AI tools themselves.

2. We do not always assume everything we read is true; we keep in mind that it might represent a biased opinion, and we utilize our common sense: Does it feel right?

3. We push back against the AI when it affirms our responses and ask it to help us understand and not to please us. Furthermore, students can interrogate the AI: when was it last trained? Does it cite sources? Could the phrasing include a hallucination?

See more steps we can take during the generate stage in my upcoming book!

Verify

Now that we have encountered information that seems to be reliable, we find its claim, references, or data points and look for an alternative source of information and verify it. Additional things that we do during this step include:

1. We deliberately slow by pausing to consider the information we are encountering, rather than trying to efficiently complete a task.

2. We analyze what we have read by using the CRAAP Test.

See more ways to verify information in my upcoming book!

Reflect

As a final step, we want to ensure that we are avoiding cognitive biases and echo chambers; the essential thing is that time is allocated and the conditions for reflective thought are right (i.e. students are not rushed at the end of class just before the bell rings, they are calm, alert). When we reflect on the overall process, we encourage metacognition which can be an extremely powerful move.

Example questions teachers can share with students to encourage reflection on biases include:

1. Default-to-Truth: What evidence or source did I find that could have made me doubt or disprove this claim?

2. AI Omniscience Bias: What might this AI not have known or been unable to do, and how did I check the information with a non-AI source?

   See more biases in my upcoming book!

Two-part series: theory first, practice second. Adults use AI for efficiency; students risk cognitive debt if they copy them. Instead of banning AI in K-12 classrooms, design learning-first workflows. To protect productive struggle, use AI to remove real barriers, and require small “breadcrumbs” of thinking plus a quick check to ensure AI supports learning rather than replacing it.

Intro

As we begin 2026 and look back at 2025, Marc Zao-Sanders’ viral HBR piece was a helpful reality check and numerous social posts focused on AI for “therapy” and “finding purpose”. Here’s another great article from the NY Times about a proliferation fo children’s toys that serve as companions as well. While, what Jason Prohaska calls Artificial Imtimacy, hasn’t quite proliferated our classrooms, I want to wrap 2025 with a topic that I see mattering most for classrooms: teachers and students are using AI to save time: drafting, planning, summarizing, organizing, polishing, creating. Most everyday K-12 use is still about efficiency.

But K-12 students can’t simply inherit adult habits under the banner of competitiveness or staying up with technology. Every adult expert earned their literacy in a fully non-AI childhood; kids in 2025 are growing up inside an experiment and the long-term impacts of AI are in education are not clear. But what I can say is that the habits we practice with kids today will become the habits they practice as adults. For students who are learning to drive their own brains, “efficiency” can replace the struggle that builds judgment and reinforce shortcuts disguised as polished work. This post argues that student AI use must be fundamentally different from adults’. We should design thinking-first, scaffolded processes that build durable strategies for problem-solving, self-knowledge, connection-making, and reasoning, and bring AI in only when it supports cognitive moves.

Humans in Wall-E vs. Iron Man

This year, I read several amazing books that influenced how I approach AI. Funnily enough, all of them are about pedagogy and learning sciences that are pre-AI. And I’m finding that I am cementing a conclusion from last year: when we understand the learning sciences, how our brains develop, and research-backed pedagogy, then we can invite AI into the classroom because it should always be a learning-first approach centered around our children. In other words, what’s best for them? How can AI support what our kids need?

One of the books I read this year was Carol Dweck’s Mindset. In this seminal book (from 2007!), she argues that the mindset that we bring into our classrooms (growth vs. fixed) can have a significant impact on our ability to learn and grow. I think that today, many people take these concepts as granted, so instead of trying to convince you of their validity, I’d like to ask, what mindset is important when we use AI? I believe there are two mindsets toward AI:

1. The assistant mindset: AI is there to complete things for me.

2. The collaborator mindset: AI is there to complete things with me.

To understand these two mindsets, there are two films that I think explore both in kid-friendly language, so please feel free to share these ideas with your learners.

In the film Wall-E, humans have become so reliant on technology to do everything for them, that they have physically atrophied. They float in chairs, distracted by screens, unable to walk when falling out of their hover chairs (see image below).

Applying this AI as an assistant mindset to the classroom, I believe that it can surface when we think of school as a series of tasks we have to complete, and the shortest pathway with the fewest bumps is the best. In that way, learners with this mindset might use AI as a tool to cognitively offload their thinking , which, when done repeatedly, can lead to inadvertent cognitive atrophy much like the humans in Wall-E who forget how to walk. And as for us teachers, we might be supporting this very mindset when we choose to, perhaps through inaction or avoiding honest dialogue about AI, design learning tasks that see the deep, effortful thinking as secondary.

Deep, slow thinking sounds great, but realistically speaking, no human is engaged in effortful thinking all the time. Deep cognition requires work and energy as one of my favorite books by Daniel Kahneman, Thinking, Fast and Slow, explains. Our brains often prefer mental shortcuts, or heuristics as he calls them, to conserve effort and energy. I am really not doing this amazing book justice, so go read it and find out more. Glossing over it, the simple takeaway for schools is this: humans are already predisposed to avoid effort in a pre-AI world, and generative AI makes bypassing effort even more accessible, and even accidentally possible.

I want to propose that we, as educators, focus on developing humans who have AI-based skills they can use as strategies to support their thinking throughout their adult lives as a form of digital literacy. I also want to propose that we support children through scaffolded processes that name the thinking and make it clear when and how AI can (or cannot) be used at different steps.

In this more collaborative approach, I believe we can construct something that looks more like Marvel’s Iron Man. The genius creator, Tony Stark, is not replaced by the AI he creates (JARVIS); he is the pilot. The AI gives him strength, and the agency remains with him to decide how to solve problems. He battles villains and flies, something no human could do before. In Wall-E, humans lose something that makes them fundamentally human, the ability to walk. With Tony Stark’s flight, he transcends what it means to be human.

This is the goal of my book, AI-Enhanced Processes (second edition currently in review for publishing on Amazon, stay tuned). We want students to collaborate with AI, not be dependent on it to function.

Classroom dialogue ideas

Do you want to be carried by the chair in Wall-E or do you want to fly like Tony Stark? How can we think of AI as a collaborator or a superpower when we use it on this assignment?

Kids Need Productive Struggle

Learning to learn, the importance of effort, and developing study skills are age-old topics, but it’s worth repeating again in this AI context: that we need to teach students how to recognize what is and is not learning, how it can feel like a struggle, and help them see that they are learning.

I would argue that a core part of AI literacy is knowing what learning feels like: being able to tell when learning is actually happening. Students need both the big picture (why we go to school) and the immediate purpose (why are we doing this task), plus a practical question: How do I know I’m making progress toward mastery (constant and ongoing feedback loops)? When students can see evidence of improvement, they can make wiser choices about when AI is helpful and when it becomes unhelpful cognitive offloading.

Cognitive offloading is tempting because our brains are constantly weighing cost versus payoff. When effort doesn’t produce detectable progress, it starts to feel wasted, and shortcuts become more appealing. Imagine what’s going through a kid’s brain when he knows AI can produce a polished result instantly.

That brings me to an idea that’s sticking with me in 2025: AI-induced cognitive debt. When AI removes the thinking in the present moment (because it’s there, and the task feels low-value) students may gain a clean output, but they pay later in weaker recall, shallow understanding, and lower confidence when they’re on their own. They “owe” the thining to the AI they used for the perceived learning. You see it when students only read summaries instead of grappling with full texts, submit writing with minimal revision (more “reading an answer” than writing to think), or copy correct equations without attempting the reasoning first.

After some brainstorming, I wanted to point out a few ways teachers could have a quick chat with students as a diagnostic to see if their students might have taken cognitive offloading to the point where they have cognitive debt.

1. Reconstruction: Could you recreate a rough version tomorrow (spaced recall) without the use of AI? If you can’t, AI might have done too much.

2. Reasoning: Can you explain why you made the key choices (claims, structure, method, etc.)? If you can’t, AI might have done too much.

3. AI Critical Awareness: What were you watching for while using AI (hallucination, weak evidence, shallow reasoning, tone mismatch, etc.)? If you can’t, you were using it AI blindly.

4. Breadcrumbs: What artifact of learning shows your thinking changed (a before/after sentence, a sketch revision, a chat screenshot with your takeaway, an earlier prototype)? If you can’t answer it, AI might have circumvented a step the process, or the process is not serving your needs. Consider adapting the process according to the learner’s age and needs. If adaptation of the process is necessary, dialogue with the learner about how they will record their learning.

When it comes to friction leading to learning, it’s also important to note that not all friction is created equally. Productive struggle is the kind that can lead to students feeling curious through a guiding question, solving a problem that they care about, exploring something they’re curious about, seeking feedback and making adjustments to their work, spaced retrieval, etc. For more on effort, read this fantastic article from Carl Hendrick, what it is, and how to help students engage in it.

What we really want to avoid is unproductive struggle, like unclear instruction, difficulty seeing the board because of seat placement, high stress levels, language barriers, boring topics, insufficient background knowledge, etc. These sorts of frictions are not stimulating or engaging and can hamper the learning experience. For more, I would recommend exploring the UDL framework on how barriers can be dropped for all learners. It’s also worth noting that productive struggle is an equity issue in many schools around the world. When we have barriers related to AI access, language proficiency, AI skill, etc., some students may actually have a privilege to overcome barriers through their use of AI. As the UDL framework encourages, we should be considering how we can remove barriers for all learners; if AI can help kids to translanguage, unpack classroom concepts, access further practice, I would wholeheartedly advocate for this approach.

Struggle is not equal to suffering or challenges in learning; productive struggle is about getting kids comfortable being in their own Vygotskian Zone of Proximal Development, and recognizing what it feels like (sometimes a little uncomfortable). Forgive another metaphor, but it’s much like going to the gym. We do not ask our personal trainers to lift the weights for us because we know, our muscles won’t grow without the effort, and we have to choose the right amount of weight that we can lift to push our muscles. Our lessons that push each learner to grow without unproductive barriers while intentionally using AI that maintains important forms of effort is the point. Teachers who address AI through dialogue with their students and establish clear processes and expectations are one way to make sure AI-use remains productive. Going back to the topic at hand, productive struggle often means taking time, and not necessarily completing a task efficiently.

Classroom dialogue ideas

How do we get really good at something; does AI help us to get better at things? Where is the effort in our work going to be placed with us as learners, and where will we think with AI? How will you know if AI is doing the work for us rather than with us? What could your effort look like in this task? How can you manage your time to make sure learning takes place?

End of Part 1. In Part 2: Practical Application

We know we want Iron Man, not Wall-E. We know we need to protect productive struggle to avoid accumulating cognitive debt. But knowing why is different from knowing how. If we stop here, we just have a philosophy, not a plan. In Part 2 of this series, I’m going to tackle the practical side: how do we move beyond the “process vs. product” debate and actually design processes that leave “breadcrumbs” of thinking? I’ll share specific strategies including the “Monday Ready” tools and an example from a classroom with a sneakpeak at the second edition of my book to help you design learning where students stay in the pilot’s seat.

Great Reads Related to This Article

These are some of the works I reference, allude to, or was inspired by while writing this article. All of them look at the ideas presented here from a different angle and are recommended reading in the form of articles or books.

• “AI Is Changing How We Learn at Work” by Lynda Gratton

• “Why Does Thinking Feel So Hard?” by Carl Hendrick

• Thinking, Fast and Slow by Daniel Kahneman

• Doing School by Denise Pope

• Mindset by Carol Dweck.

AI Disclosure

Parts 2 of this article was created through an iterative Think, Generate, Edit process, repeated multiple times. The initial draft was written by me and informed by my work on AI-Enhanced Processes, Second Edition. I then used multiple AI models: ChatGPT, Google Gemini, Claude Sonnet, and ChatGPT Deep Research. They were used to refine, challenge, and extend my thinking. Visual elements were generated with ChatGPT. All writing and final decisions were reviewed and edited by me, a real human, to ensure coherence, intent, and accuracy.

It’s also worth disclosing how long it took me to write this article. I would estimate 8-10 hours were spent writing, thinking, rewriting, generating, etc. I would not say that was a particularly fast way of working, rather AI was my collaborator.