AI Can't See Your Blind Spots. Neither Can You.

Cities of Circles 12 × 12

AI can process everything you tell it. It cannot see the thing you don't know you're not telling it.

3 minute read

Over the last few months, I've written about where AI actually creates value and what it takes to deploy it without creating a mess. All of that still holds. But there's a category of risk none of it touches. It has nothing to do with the technology. It's you.

By definition, a blind spot is something you cannot see. That's not always a failure of effort or intelligence. Sometimes it's motivated: certain patterns operate outside your awareness because looking at them directly would cost you something, comfort, identity, a version of yourself you'd rather not examine too closely. And sometimes it's just habit: a pattern repeated so many times it stopped registering as a pattern at all, not defended, just automatic. Either way, you cannot think your way around it, because the thinking itself is shaped by it.

AI does not fix this. It cannot, because AI only ever sees what you show it: your framing, your inputs, the question you decided to ask. If the way you've framed a problem is already distorted by a pattern you can't see, AI will not catch that. It will optimize beautifully within the distortion and hand you a confident, well-reasoned answer to the wrong question.

The motivated kind

A leader who avoids conflict doesn't usually ask AI, "should I even be having this conversation?" They ask, "what's the most diplomatic way to say this?" And here's the part that matters: they are not hiding anything when they ask it. As far as they can tell, diplomacy is genuinely what the moment calls for. The avoidance doesn't feel like avoidance from the inside. It feels like patience, or good timing, or simply being thoughtful about how something lands.

That's what makes it motivated rather than a decision anyone actually weighed. Ask this leader directly whether they avoid conflict, and they might even agree. What they won't do is connect that admission to the message they just asked AI to help soften, because the two never get filed under the same heading. AI will draft something warm, careful, and well-constructed, and it will make that leader more efficient at avoiding the exact conversation they needed to have, without either of them ever framing it that way. The tool did its job perfectly. It just did the wrong job, because the real question, whether this conversation needed to happen at all, was never actually on the table.

The habitual kind

A leader I know has opened every strategy conversation with a SWOT analysis for fifteen years. Not because it's still the sharpest tool for the problem in front of her, but because that's simply the shape a strategy conversation takes in her mind. Ask her directly and she'd readily admit it's a habit, no defensiveness required. What she's never done is ask whether the habit still earns its place.

That's the quieter risk. She'll ask AI to "build out the SWOT for this," and AI will produce an excellent one. It won't ask whether SWOT was the right frame to begin with, because from the outside, a fifteen-year habit looks exactly like a considered choice. Nothing about the request signals that it's on autopilot.

This is the paradox worth sitting with: the better the tool gets, the more convincingly it will help you execute a pattern you can't see, whether that pattern is defended or just worn smooth by repetition. Either way, it arrived looking like a reasonable request.

What this means for how you use it

I'm not telling you to use AI less. I'm telling you to know your own patterns before you hand a tool the job of executing them faster. The prompt was never the risk. What was underneath it, that's the risk.

Technical fluency won't stay rare for long. Everyone will have it within a couple of years. Self-awareness won't catch up the same way, because no model reaches it. That's the actual asset here, and it's one you have to build the old-fashioned way, by looking at yourself with someone who won't let you off the hook.