A 20/80 rule for AI in education

And what that looked like yesterday in my classroom

Yesterday my students used AI for 10 minutes. That was more than enough, in both senses.

I teach a class for 11th and 12th graders focused on purpose development and career-connected learning. Each period is about 45 minutes long. On the days we use AI tools, we use them for about 10 minutes each time.

That’s roughly 20% of our class that’s focused on AI.

The other 80% is focused on projects, reflection, discussion, and exploration.

I’ve found that to be the right ratio to balance familiarity with AI and familiarity with the many other concepts and skills we still need to teach.

That ratio also helps to ensure students are the focus 100% of the time.

Here’s what I mean:

Background for yesterday’s lesson

The first two days this week, students launched individual projects related to the impact they hope to have on the world and the career that they feel will help them have that impact. (They got help designing those projects from an AI tool I built for that purpose, which they used for about 10 minutes last Friday.)

Among many other things, students are writing a novel about high school, coding an app to design cars, opening a small event-planning + baking business, exploring wildlife veterinary initiatives in our area, and launching a podcast about the mental element of athlete’s recovery from injuries that will feature both student athletes and professional athletic trainers.

Walden, ikigai, RIASEC

Yesterday we stepped back to look at those projects, and students’ experience of them, from a distance, to enable a metacognitive vantage point.

To make our metacognitive undertaking clearer, I opened our 45 minute period with a favorite line from Walden: “With thinking we may be beside ourselves.”

I then said, “The point of our projects is not really our projects. They’re important of course, and I’m excited to see where you all take them. But the real point of your project is you. Your project gives you an experience to reflect on. It’s a mirror to help you see yourself from a new angle. You’re the real project.”

Next students reflected on their project experience so far through the lens of the “ikigai” framework, which is an anchor for our course.

Students wrote quietly in their notebooks as they considered what part of their project they felt good at, what part they most enjoyed doing, and what needs in the world their project was addressing.

After they did that, students took a short, student-friendly version of the RIASEC survey on paper to determine their personality types and the careers that align most with those types.

Then it was time for AI.

Prelude to a bot

Before we opened the new bot I had built for class yesterday, I gave this rationale:

Accenture, a big consulting firm, announced last week it was firing thousands of people who weren't able to learn to use AI. [A few students each period gasped audibly.] We need to make sure you learn how to use AI and get used to these tools now, while you’re still in high school. Not because you should be using them all the time (you shouldn’t), or because you need to worry about your future job (you won't precisely because we’re doing things like this). But we need to make sure you’re getting familiar with how AI works, and what the right ways to use it are. Today I reprogrammed our robot to help us with our project reflections¹, but later this year I’ll be showing you how to do the programming yourself.

AI in action

Students then individually opened the bot, which guided them through a series of questions. The first four questions were the same for everyone, based on the survey they just completed and the ikigai reflections they did before that:

1. Hello. I’m here to help you reflect on your project and your recent RIASEC results. Let’s see if we can uncover some interesting connections about your path forward. First, could you please share your top three letters from the RIASEC test?

2. Now, let’s connect this to your current career-connected project. First, what has been a part of the project that you’ve felt you were genuinely good at?

3. And what is the part of the project you’ve truly enjoyed, the part that’s felt energizing and engaging?

4. Now, thinking bigger, what part of your project addresses something you believe the world needs?

It then asked two follow-up questions to help students reflect a little more deeply and learn a little more about how they’re thinking about their career and future.

Finally, it ended by giving them a metaphorical title that tried to capture the current spirit of their career and purpose journey.

For example, it gave the aspiring trainer and podcaster this:

Right now, you’re approaching life like a lantern bearer, using conversation to illuminate the often-hidden mental paths of recovery so that others don’t have to walk them in the dark.

Students thoroughly enjoyed these. Some others included:

• Cartographer of the mind (for an aspiring neurologist)

• Builder of legal fortresses (for an aspiring community-focused attorney)

• River captain on a fast-moving current (for an aspiring investment advisor)

When students got less fun sobriquets, I told them to tell the AI, “Make it more dramatic.” One aspiring police officer who typed that in went from “community protector” to “frontline guardian”, and an aspiring entrepreneur went from “brand builder” to “architect of a business multiverse.”

I myself got excited enough about the effect these metaphors seemed to be having on students that I suggested they might want to use them as Instagram handles. This went over even more poorly than my dad jokes do.

End scene

After that we turned off the AI and put away the screens. Before we moved on, I shared this closing thought:

You see how some people didn’t like their metaphor and asked the AI to make it more dramatic? That’s one of the key AI skills many people still don’t have. It’s a machine, a tool, that works for us. If you’re using a screwdriver and it doesn’t fit the screw, you get a new top or find another screwdriver. That’s what AI use is in some ways: going back and forth until it gives you what you’re looking for. That’s the kind of skill that’s increasingly essential in every career field you all are exploring. Podcasters and entrepreneurs and doctors and designers will all need to know how to do that, to iterate with AI.

And then you see how we turned AI off when it gave us what we needed, when you each got your metaphor? That to me is another key AI skill — knowing when you don’t need it any more. Once the screw is in place, you don’t carry the screwdriver around all day. Now let’s be clear, AI is a very useful screwdriver. As you can see, I use it a lot for our course. Maybe a Swiss Army knife is a better metaphor for it. But however we think about it, we need to remember AI is a tool we can learn how to pick up and when to put down.

We then ended class with a final handwritten reflection, on students’ thoughts about their metaphor, their project and their evolving sense of purpose and career.

In other words, the end of class was about the students, their learning, and their lives. So was the start. AI only showed up in the middle, said its few lines, and exited stage left (or screen down, as it were). Students were the stars.

AI metacognition

The next sentence after that earlier Walden quote goes like this: “By a conscious effort of the mind we can stand aloof…and all things, good and bad, go by us like a torrent.”

The torrent of AI development, AI hype, and AI doomerism can make it feel like school as we know it is being swept away, leaving us scrambling to find our footing.

But by conscious effort of the mind we can stand aloof from AI. Not in the luddite sense. In the sense of thinking, discussing, and deciding when, how, and when not to use AI. In the sense of considering when to put our boats in the water and oars in the current, and when to walk together along the shore.

And, as usual, our conscious efforts will help students make their own.

1. For those interested in prompting nitty-gritty:

I had Gemini 2.5 Pro create the bot’s prompt for me, using this prompt to get Gemini started:

You’re an expert in LLM prompting, career-connected learning, purpose development, and the RIASEC framework. Students in my 11th and 12th purpose and career class are taking a quick RIASEC test tomorrow and then I want an LLM bot to help them reflect on their current sense of ikigai based on their current career-connected project and their RIASEC to reflect on their life journey ahead after high school. For the ikigai portion of class, they will ask themselves:

What part of my career connected project do I feel good at?

What part of my project do I enjoy?

What part of my project does the world need?

What career will pay me to do this kind of work?

So the bot needs to take those reflections, the results of the attached RIASEC test, and then help students reflect for a few minutes. This reflection should be based on all relevant best practice and research for students this age doing this kind of work and thinking. The total conversation should be 5 questions that then roll up into a summary of the student’s own thinking, so they can see themselves from the outside and hopefully discover something new about themself.

You’re about to write the prompt. It’s a prompt for a bot using Gemini 2.5 Pro so tailor appropriately for that model’s underlying design. Before you write the prompt though ask me any clarifying questions it would be helpful to have answered to make the prompt as effective as possible.

Gemini asked me four clarifying questions to help it align the bot's prompt as much as possible to my goals; I answered those questions in this second prompt:

1. I want the bot to be a wise mentor.

2. Look for a subtle bridge between areas of their personality but also point it out to students so they can reflect.

3. I want the summary to give them a view of themself from a new angle. Like a wise mentor helping them understand themselves better. It should leave them with a metaphor or image that captures their approach to life and career right now. “Right now you’re approaching life like a....” or “Right now you’re seeing life like a...”

4. The bot should ask for their RIASEC code and their ikigai reflections one by one. So one question for RIASEC and three for ikigai, one by one. Then a couple of follow up questions to dig deeper. Then the final roll-up and metaphor/image.

It then wrote the full prompt I used to create the bot for class.