leading through the ai transition

The hardest part of running a team right now isn’t deciding which AI tools to adopt. It’s helping engineers re-anchor their identity when the work itself is changing under their feet.

Most of the conversations I’ve had this quarter weren’t about productivity. They were about meaning. When the bottleneck moves from typing to deciding, the part of the job people loved often vanishes — and the part they tolerated becomes the whole job.

What’s actually changing

The framing that’s helped: stop calling this a productivity transition. Call it a craft transition. The skills that compound from here aren’t typing speed. They’re judgment, taste, and the ability to direct a system that’s faster than you are.

A small example. Last month I asked an engineer to write a tool. Pre-2024, that meant a week of work. This time it took ninety minutes. The engineer was uneasy — not because the tool was bad (it wasn’t), but because the part of the work they were proud of had been compressed into something that felt thin.

// the ninety-minute tool
export function summarize(input: string): string {
  return input.split('\n').slice(0, 3).join('\n');
}

The conversation we had next is the one I think every engineering leader is having or about to have: what is your craft now?

Three things I’m trying

1. Reframing reviews. Less “did you write this code?” — more “did this code do what it needed to?”
2. Making space for taste. Code review now spends more time on architecture, naming, and whether the abstraction earns its weight.
3. Holding the line on shipping. Speed without judgment is just faster failure. We’re slower than we could be on purpose.

None of this is novel. But the rate at which it has to happen — and the emotional weight underneath it — is unprecedented in any tech transition I’ve lived through.

More to come on this.