When you don’t know what you don’t know

Sometimes you're pulled into something so far outside your experience that you don't even know what a competent person should ask. The stakes are real, the timeline is short, and you're aware that there's a risk in missing something that would be obvious to people who've been here before.

There's a particular discomfort in these moments. You're expected to engage with confidence, even though you don't yet have a reliable sense of what matters, what's minor, and what could create problems later.

Here's the thing about the people who have been here before: what we call their experience is largely a library of failure patterns. They've watched this kind of project stall, seen this kind of contract go sideways, sat in the postmortem for this exact mistake. That's why they ask better questions — not because they're smarter, but because every question is backed by a memory of what happened when nobody asked it.

You can't borrow their judgment. But you can borrow their pattern library.

In these moments, I don't use AI to get answers. I use it as a stand-in consultant — someone who has seen a thousand versions of the situation I'm seeing for the first time. I ask:

• "What are the common failure modes in situations like this?"
• "What do experienced people tend to worry about first?"
• "What questions usually get asked too late?"
• "What looks low-risk on the surface but often isn't?"

Notice what these questions have in common: they're the questions experienced people carry without knowing they carry them. Nobody taught them to ask "what gets asked too late?" — they just remember the time it was them, asking it too late.

The goal isn't to sound smart or to appear decisive. It's the opposite — to avoid false confidence long enough to engage responsibly. Borrowed pattern recognition doesn't tell me what to decide. It tells me where the terrain drops off: where caution is warranted, where real expertise is needed, and where an early offhand signal could set a direction nobody intended.

That last one matters more than it sounds. In unfamiliar territory, the biggest risk usually isn't the decision you get wrong. It's the casual comment you make in week one that hardens into an expectation by week three.

So the value of AI here isn't speed, and it definitely isn't answers. It's that it keeps me in the conversation long enough to learn — asking the questions a veteran would ask, flagging where I need to pull in the people who actually carry the scars — instead of either faking fluency or going quiet.

Because in moments like this, the race isn't against the deadline. It's against the point where confidence outpaces understanding. Borrowed pattern recognition is how you stay on the right side of that line.