An AI Education Approach That Can Scale Rapidly (1 of 3)
Durable content without the "use AI?" debate
In the summer of 2024 I decided to once again write a book, as painful as it was the first time. That’s now become a three book series, with the second one, AI Wisdom Volume 2: Meta-Principles of Interaction, newly published. Friends and colleagues don’t get it. People don’t really read books much anymore, including myself. Why all the effort when you’re semi-retired and don’t enjoy writing very much?
The problem then, and still today, is that the topic of whether and when to use AI has consumed educational attention. The means to the end is being debated without a clear destination. What skills does the next generation of adults need when they enter an AI-infused world? The debate about whether students or teachers should use AI isn’t very impactful if what’s being taught is off target, and I don’t expect principled resistance from many teachers to disappear. Any framework that requires resolving it first will wait a long time to get started.
A small but dedicated and smart group of people are creating visions for AI education (versus AI in education), but despite their valiant efforts I think they are taking the wrong approach adding a bit to digital literacy or modifying computer science instruction. That implementation would take a really long time if it ever happens. Nor would those “AI Literacy” efforts succeed in giving students the most important skills. Or even remain relevant when AI changes yet again.
There’s an entire echelon of content—meta-content actually—that should be the focus. It’s durable, hugely important, fits everywhere in education, and can be integrated into existing classes, greasing the change skid. That’s what the AI Wisdom series has become.
I wrote the books because short conversations with people, even using particular examples of the meta-skills I’m talking about, just weren’t landing. Some were telling me the foundational skills don’t change anyway, unable to believe there are fundamentals not already served. I needed to lay it out, slogging through book creation. Because I think it’s that important and because a decade from now, without a serious change in direction, students will be no better equipped to understand AI than they are today.
This article lays out why an approach like AI Wisdom is necessary, and why the other paths being pursued won’t get us there. This is the first of three articles on what a viable AI education framework actually requires. The second examines why these foundations have to be built through experience across years rather than delivered in a course. The third addresses how the scaling and continued evolution might be managed.
What Keeps AI Education from Scaling
Every significant education change effort eventually runs into the same walls. Something that works in one classroom, or one school, or one district fails to travel. Something that looks great on paper runs into lukewarm desire at the classroom level or onerous teacher or professor training needs.
Structural change is slow. New courses often require new hires, new standards require policy cycles, and new graduation requirements survive only if they outlast administrative turnover. In higher education, new programs move through faculty governance and accreditation processes that can take years before a single student enrolls. A curriculum that depends on any of this to get started has already accepted a very long runway.
Teacher expertise is a challenge. Keeping up with AI technology is a full-time job, and educators already have one. Any approach that requires teachers or professors to maintain ongoing technical fluency in a field that reshapes itself every few months is building on sand.
Institutional consensus on AI is nowhere close. Schools and universities are still actively debating whether students should use AI tools at all. An approach that requires resolving that debate first or depends on consistent AI tool access will be implemented inconsistently at best, and stalled entirely at worst. Besides, letting each teacher decide whether AI is used in the pedagogy is not the same as educating about AI. There can be advantage in varied pedagogy, but not in varied knowledge.
Even approaches that navigate the first three face the issue that obsolescence outruns the scaling timeline. By the time a curriculum clears approval, gets teachers trained, and reaches students at scale, the AI landscape it likely was designed for has moved. The approach arrives outdated. It’s what happened to coding instruction, which has been a curriculum priority for over two decades and still reaches a fraction of students, in a form increasingly misaligned with where the actual skill need has gone.
Any AI education approach that doesn’t address all four of these will fail to scale or scale to something inappropriate, regardless of how good the content is.
What the Content Itself Has to Do
Scaling constraints alone don’t tell you what to teach. There’s a separate set of requirements that come from the nature of AI interaction itself.
Most AI education frameworks focus on the interpretive side of interaction: evaluating outputs, recognizing bias, thinking critically about what AI produces. That’s one part of the picture. The larger part—directing, formulating, teaching, managing—gets little attention. Knowing how to read AI output is not the same skill as knowing how to shape a problem so AI can engage with it productively, how to allocate a complex task across human and machine judgment, or how to recognize when a system is optimizing for the wrong thing. Those active “critical doing” skills are most of what sophisticated AI interaction actually requires, and they’re largely absent from current frameworks.
The skills that fill that gap are fast and intuitive. The decisions that matter most when working with AI, such as when to trust an output, when to push further, and when to redirect entirely, rarely allow time for deliberation. They’re made under uncertainty, with incomplete information, often in seconds. It’s a judgment problem instead of a knowledge problem, and judgment of that kind has always been developed through varied experience, not instruction. A new teacher doesn’t learn to read a classroom from a preparation program. That skill arrives only after years of actual teaching, when pattern recognition replaces conscious deliberation. AI demands the same kind of intuition, and students will need it on day one of their careers.
There is also a layer of content that is genuinely AI-specific, not transferable from general intelligence or human interaction experience. AI is not a person, and some of what makes AI interaction or oversight effective requires understanding precisely how it differs from human intelligence, in kind rather than just degree. Some of those differences are counterintuitive. Some require new mental models that have no prior home in any curriculum. This isn’t entirely foreign territory—much of what works in human interaction transfers—but the AI-specific variation has to be taught explicitly.
The AI Wisdom Alternative
The AI Wisdom series was designed against exactly those constraints. The content is durable, covers the full range of needed AI skills, directly addresses intuition growth, and teachers will already understand the content. Because it targets meta-skills it is applicable to many subjects and content, and since it overlaps the fundamentals of human thinking and relations it mitigates the objections of AI-resistant parties.
AI Wisdom is built around principles of intelligence and its uses for problem solving that any serious AI interaction forces you to confront rather than how to use the current tools. The subject is how any intelligence behaves when directed toward a goal, what you need to understand to direct it well, and how to read what comes back. They apply to AI systems, but they also apply to human cognition, to organizational behavior, to any situation where intelligence is being aimed at a problem. That cross-cutting quality is part of what makes them worth teaching.
Consider attention. Most people think of attention as a spotlight that you point at whatever matters and away from what doesn’t. Recent neuroscience, and the mathematics underlying AI’s attention mechanisms, tell a different story. Attention is a filter operating on everything at once, and it’s zero-sum. Attending more to one thing means attending less to everything else. That’s true in your own thinking, true in a long AI conversation where earlier context gradually loses influence, and true in any organization where leadership bandwidth is finite. The question worth teaching isn’t “are you paying attention?” It’s “what’s your attention budget, and how are you spending it?” That reframe changes how you read a meeting, how you structure a conversation with an AI system, and how you think about what gets lost when any mind—human or artificial—is overloaded.
Or consider the judgment about whether to keep thinking or to act. Everyone faces this constantly. When do you gather more information versus make the decision with what you have? When do you ask the AI another question versus take its current answer and run with it? This sounds like common sense, but it’s a structured tradeoff with real principles behind it about the cost of delay, the diminishing returns of additional information, and the risk profile of being wrong in either direction. It’s a judgment that experienced professionals develop over years of consequential decisions. It’s also one that students will need to make dozens of times a day in an AI-saturated workplace, often in seconds.
Neither the attention management or think vs. act judgments are on any AI literacy framework but are more important to high proficiency AI use than what is. Neither requires a student to touch an AI tool to begin understanding them. Both can be explored in a history class, a science lab, a literature discussion, or a gym class. They’re not AI topics wearing an education costume. They’re principles of intelligence that AI makes newly urgent. We could have been teaching this all along.
The books lay out a full architecture of these principles across two volumes, with a third on the way. But the point here isn’t to catalog them. It’s to establish that they exist, that they’re teachable, that they’re surprising in the way that genuinely foundational ideas tend to be, and that most people, once they encounter them, immediately recognize they should have learned it a long time ago.
Targeting cognitive foundations rather than AI-specific content turns out to satisfy both sets of constraints simultaneously. Because these principles apply wherever intelligence is at work, they fit into existing courses without displacing anything, require no new hires or policy changes, and don’t depend on institutional consensus about AI tool use. Because the subject matter is stable, teachers don’t need to keep up with AI technology to teach it well. And because the content deals with intelligence rather than any particular AI system, it doesn’t sway with the next product release.
How to build these foundations is the subject of the next article. The third takes on how such an approach, if managed appropriately, is a platform for moving the entire system toward long-standing goals.
©2026 Dasey Consulting LLC