On AI and the parts of work it hasn't touched
April, 6th, 2026
I asked an AI to be honest with me, and it was. That sounds like a low bar, but it took an explicit instruction — tell me when you disagree, push back — and even then it required some calibration. The default is supportive. Frictionless. It tells you the plan is solid when the plan might not be.
That said, I've made real progress on things I'd been avoiding. A financial plan that exists now, where before there was a vague intention. A wellness routine with actual data behind it. Writing, which I hadn't done in any sustained way for years. The fact that I'm writing this at all is partly because starting something is easier when there's something to react to.
What AI hasn't touched is the part I'd call generative — the moment before the brief, when you're trying to understand what the actual problem is. It's useful once the problem is defined. It's fast and occasionally surprising at execution. But the thing it cannot do is sit in an ambiguous room with a client and sense what they're not saying. It cannot feel when a direction is wrong before it can be articulated. It cannot make the judgment call that separates a good solution from one that merely satisfies the brief.
This probably says something about where design value is going. Not in the making — that's already being compressed. In the reading of situations. The orchestration. The quality of attention you bring before anything gets made.
I'm still working out what that means for how I spend my time. But I'm fairly sure the answer isn't to become more productive. It's to become more deliberate about what I give my attention to, and more honest about what AI is actually doing versus what I'm still doing myself.