The AI Framework I Shared with My Kids (and Why It’s for All of Us)

I’ve been thinking a lot about how my kids are going to use AI at school. Education really feels like the front line of this shift, and while I can’t—and wouldn’t—force them to use these tools in a specific way, we’ve been talking through a framework that I’ve found essential in my own work.

They seem to get it, and I’m hoping they’ll trust that this is the approach that actually sets them up for success. Because I’ve realized that when people struggle with AI, or feel like it’s just not “working” for them, it usually comes down to skipping one of these steps.

The Framework at a Glance

This is the logic I’ve shared with them. It’s a filter to run through before you even start:

1. Do you know what you’re asking for? If no, pause and explore.
2. Can you tell if the answer is good or bad? If no, you are in the learning loop.

The Win Condition: If you can do both, you have a 10x multiplier.

1. The Mirror: Do You Know What You’re Asking For?

I often tell my kids that AI is a mirror. It reflects exactly what you put into it. If you provide lazy, low-quality input, you’re going to get a lazy, low-quality output. It will do the bare minimum because you did the bare minimum.

The reason concepts like “Prompt Engineering,” “Vibe Coding,” and “Spec Coding” have become such massive topics of conversation is actually rooted in our inherent nature to ask really bad questions. As humans, we are naturally vague; we rely on shared context and “vibes” to communicate. But when we treat AI that way, it fails us.

“Vibe coding”—where you essentially throw a general feeling at the AI and hope it builds what’s in your head—is the starting point for many. But the people who succeed are the ones who move toward “Spec Coding”. These aren’t just technical buzzwords; they are techniques developed specifically to solve the problem of our own lack of clarity. They exist because, without a rigorous process, we almost always fail at Step One.

The challenge in schools is that it’s now very easy to put in low-quality effort and get something back that looks like a great result. But it’s a reflection with no depth.

For example:

The Lazy Question: “Write me an essay about the Roman Empire.”

Result: You get a finished product you didn’t earn. You didn’t learn anything, and you can’t defend a single point in it.

The Learning Question: “I need to write an essay about the fall of the Roman Republic. Can you help me understand the three most important political shifts that led to it, and explain why historians still debate them?”

Result: Now you’re using AI as a tutor. You still have to write the essay, but you walk in with a real understanding of the material. That’s the difference between a shortcut and a head start.

If you don’t know what you’re asking for yet, that’s okay. Sometimes you have to ask a few “bad” questions just to find the right one. But you have to be willing to put in the thought to find that clarity. Learning is not negotiable.

2. Can You Tell if the Answer is Good or Bad?

This is the fundamental question of quality control. If you don’t have the ability to understand if the answer coming back is right or wrong, you’re blindly following a tool that might be leading you down the wrong path.

I see the difference in my own work every day:

• Low confidence — The Legal Agent: I have a trained agent for legal research for my agency. Since I’m not a lawyer, I can’t fully qualify the work. I have to use it with a high level of caution because I can’t be 100% sure it’s in my best interest.
• High confidence — The Product Manager Agent: Product management has been my career. Here, I know what “good” looks like. I can spot a flaw in the syntax or logic immediately and redirect the tool.

That second scenario—where you have the domain knowledge to verify the quality—is the gold standard.

But what happens when you’re still building that domain knowledge? That’s where the Learning Loop comes in.

3. The Learning Loop: Learning is Non-Negotiable

This is where the framework really plays out at the dinner table. When my daughter uses AI, I want her to stay in the Learning Loop.

If she doesn’t know what a “good” response looks like yet, she shouldn’t just take the answer and turn it in. That’s just cheating yourself out of a skill. Instead, she stays in the loop. She asks: “How did you get that answer?” “Why did you choose this conclusion?”

It’s an iterative process of digging into the logic until the knowledge becomes her own. Once she understands the why, she can qualify future responses herself. At that point, she’s moved from being a “student” of the tool to a “master” of the craft.

The Education Front Line: Moving Beyond Suppression

We have to be honest about where education is going. Trying to suppress AI in schools is a losing battle. You can’t stop the future, and you certainly can’t confidently determine if an essay was written by a human or a machine anymore. Detection is a dead end.

The winning strategy for schools (and parents) is figuring out how to engage in a way where the “work” isn’t just the output, but the thinking behind it.

• AI as a Tutor, not a Cheat Code: If you ask AI to do the work for you, you’re cheating yourself. If you use it as a mentor, a sounding board, or a teaching assistant, you’re learning.
• Skills over Syntax: Just like coding syntax is becoming less important than the logic behind the code, punctuation and essay structure are becoming less important than the ability to synthesize ideas.

We need to shift the conversation toward the skills we actually want to impart. We have to support students in a way that respects the power of the tool while acknowledging that the human in the driver’s seat still needs to know how to drive.

Whether you’re a student or a professional, the goal is the same: stay in the driver’s seat. Use the loops to learn, and once you’ve mastered the craft, use the mirror to win.