What I'd Do With 30 Minutes a Day of AI — A Letter to the Teacher I Was

1. To the busy-teacher version of me

If you have ChatGPT and you only tried it for a week or two, you probably don't get what all the fuss is about. You asked it some questions. You got some reasonable answers. You drafted a parent email. You sketched a lesson. Then you moved on. That was me eighteen months ago.

Here is what I would tell that version of myself, the busy-teacher version with thirty minutes a day, about how to actually use AI.

The short version is this. Most of what you tried in that first week was the lowest-leverage use of the tool. The interface you were stuck inside cannot do the work that actually saves a teacher time. And the framing in your head, that AI gives a good teacher their evenings back, is not quite right either. The real promise is choice over what to do with the time AI helps you reclaim.

2. The diagnosis: opinion-asking is the lowest-leverage use

Let's start with what I did wrong. I asked AI to make me a student-centered lesson on a topic I already had ideas for. ChatGPT made its version. The pacing was off. The concepts were too simplistic. Not all the learning objectives were covered. I had already done the thinking and had everything I wanted in my head. I just couldn't quite formalize it so I just handed the design over and hoped AI would write it up better than I could.

That doesn't work. I should have just told it exactly what I needed and got it to find the resources for me, instead of trusting it to dictate what happens in my classroom. I should've put thought into it before I wrote any prompts.

According to Education Week's research center, teacher AI use jumped from 34 percent in 2023 to 61 percent in 2025. The most common uses are lesson planning, grading support, generating materials, and drafting parent communications. Almost all of that activity sits inside a chat window. RAND's nationally representative survey of US teachers found similar patterns. About a quarter of English, math, and science teachers used AI for instructional planning in 2023 to 2024. Only 18 percent of principals reported that their school had given guidance on AI use.

So a lot of teachers are using AI. Most of them are using it in the same chat-window mode I was. And the research on what that mode actually does for learning is sobering. Fan and colleagues ran a randomized trial with 117 university students. The students who used ChatGPT wrote better essays. Their knowledge gain and their ability to transfer what they had learned, however, were not significantly different from the control group. The authors call this metacognitive laziness. Once again, AI generated outputs may look more well put together, but that doesn't mean the thinking underneath is anymore. If we as teachers are also using AI like this, not only are we metacognitively lazy, we also fail to acquire the AI skills we should be teaching our students.

If you outsource the design choices to AI, the lesson plan looks polished. The teaching does not get better. The teacher does not get better.

Ironically, most teachers know that students shouldn't use AI to do their work, and that over time, students relying on AI for the wrong things will harm their ability to think. Yet, as teachers, we do the exact same thing in pursuit of efficiency.

3. The tier above chat

The second mistake I made was assuming chat was all there was. It is not.

There is a tier above chat. Ethan Mollick writes about the shift from chatbots to agentic AI systems that can complete hours of knowledge work without the user coordinating each step. The point is the work, not the conversation. You give the tool a goal and a set of files. It runs. It produces work product. You curate, give feedback, and the AI fixes things up until you're happy.

For me the concrete version of this is Claude cowork. ChatGPT has its own version of this with custom GPTs and Operator, though file work has been weaker in my experience. AI that can sit with your files, run multi-step jobs, and hand back a finished thing is not the future. It's already here, and while some teachers are rewriting whole textbooks into HTML files that can toggle between Lexile Levels, most teachers still don't know what to do with the capability.

Teachers are starting to learn, though. Sarah Sparks reported in Education Week on a national US training program that is moving 400,000 teachers from basic chat use to building their own agentic tools. One Brooklyn teacher she quoted talked about the difference between a tool that can write him a lesson plan and one that can stress-test his lessons for content gaps and confusing wording. Six in ten teachers use AI, but most of them use it for basic lesson plans and admin work.

4. Two productive modes

There are two productive modes for teachers using AI. The first is having AI build a tool to your specification. The second is having AI generate resources you then curate. The work product is the point. The conversation is a means to the work product.

For example, in one working session, I got AI to do 3 jobs in succession. I had Claude split my AS Economics textbook into 53 chapter files. That took under 15 minutes. I then had Claude sort my MCQ bank, which holds over a thousand questions, by learning objective. Every question came back with a correct objective attached on the first pass. I then had Claude take the 53 chapter files and the labeled MCQ bank, build 53 lesson decks (one per chapter), and insert formative checks at the right points in each deck, drawing the questions from the matching learning objectives in the bank. The chapter split fed the deck structure. The labeled bank fed the formative checks.

I went to sleep after getting Claude started, and woke up tomorrow with everything finished. I could not have done any of that in a chat window. Fifty files is too much for chat. A thousand-question bank is too much for chat.

Once that anchor is in place, I got even more specific tasks done. I had Claude generate chapter-specific writing assignments for 12 units of my course, with answer keys produced alongside. Other teachers are working in this productive mode too. The 74 reported on two US teachers, Janice Donaghy and Jean D'Aurio, who used AI to build a full lesson on ancient Greek vases. Their tools were Canva and Diffit, not agentic AI in the cowork sense but the pattern is the same. Specify what you need. Have the tool produce a usable artifact. Curate it for your students. D'Aurio said classroom preparation goes from hours to seconds.

5. The 30-minute cadence

If I only had 30 minutes, I think there are two categories that pay off. The first is AI use that saves teacher time long-term. Splitting a textbook, organizing an MCQ bank, and building formative-check decks all sit here. The second is AI use that improves student learning. This is where you generate differentiated readings, build a sparring-partner exercise around a concept students always get wrong, or curate an AI-marked low-stakes practice set students do before the real assessment.

Alternate the two by week or by day depending on what your teaching needs. Some weeks the bottleneck is your preparation time. Some weeks the bottleneck is what is happening in the room. Spend the 30 minutes where the bottleneck is.

6. The ceiling is much higher than this

I also had Claude help me build app.reedlet.com, an autograder that lets teachers create custom chatbots their students use to learn. The same tool that built the 53 decks can also build a working app. It wasn't a smooth process, and the app is nowhere near perfect, but I learned a lot making it. The point is that the leverage on AI tools goes much higher than what you and I both currently know. Building with AI is both a creative and scientific endeavor.

7. What AI actually gives you back

I don't want to say that 30 minutes a day of the right AI use will give you your evenings back. That is not what I have found.

What I have found is that AI gives me choice over what to do with the time I reclaim. The textbook split saved me hours of preparation. The formative-check decks saved me weeks of small repeated edits. The thing is, though, that work would've been so onerous to do without AI that I probably wouldn't have done it to begin with. Effectively, AI made doing difficult work easier, and as a result, I think I am working more. I'm sure, though, if you wanted AI to only improve your efficiency, it could.

But where's the fun in that when you have access to a tool that is only limited by your imagination?

8. A note for the head of department reading this

If you are a middle or senior leader at a school where most of your teachers are still in the chat-only tier, consider PD that puts teachers on the tier above chat for one or two real jobs they already do. Splitting a textbook. Organizing an MCQ bank. Building one formative-check deck. The leverage is in the work product the teacher walks out with at the end of the session. The teacher who can specify a tool and curate a resource is doing different work from the teacher who is still asking AI what it thinks.

A note on method: this issue was produced through the co-creation workflow I'm advocating. The idea, the angle, the practitioner observations, the curated sources, and the final wording are mine. An AI assistant calibrated to my voice (through a guide of phrases I've approved and rejected) did the research legwork on sources I selected and drafted from an outline we agreed on.

References

1. Fan, Y., Tang, L., Le, H., et al. Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology 56(2), 489–530. 2024.
2. Kaufman, J. H., Woo, A., Eagan, J., Lee, S., & Kassan, E. B. Uneven Adoption of Artificial Intelligence Tools Among U.S. Teachers and Principals in the 2023–2024 School Year. RAND Corporation, RR-A134-25. February 2025.
3. Langreo, Lauraine. More Teachers Are Using AI in Their Classrooms. Here's Why. Education Week, January 2026.
4. Mollick, Ethan. Co-Intelligence: Living and Working with AI. Portfolio / Penguin Random House, 2024.
5. Sparks, Sarah D. Teachers Move Beyond AI Basics to More Sophisticated Instructional Uses. Education Week, 20 March 2026.
6. The 74. Case Study: How 2 Teachers Use AI Behind the Scenes to Build Lessons & Save Time. April 2024.