The interesting parts of the job are gone

The AI took over the research, writing, and analysis. What's left for the person is checking, formatting, and the odd hard case. The strongest people lose interest or leave, and the organization loses the judgment that came with them.

• People describe their job as editor, reviewer, or babysitter.
• Effort drops in the roles the AI changed most.
• Strong performers are leaving those roles.

Who: two strong performers in the roles the AI changed most, asked privately. Ask: "What share of your week is the part of the work you'd call your craft? What was it a year ago?" Look at: who has left those roles in the last six months. You're probably looking at this pattern if the craft share dropped and strong people are leaving. Probably not if the craft share held or grew.

Checking can't keep up. That pattern is about the volume of review. This pattern is about what the job has turned into.

No move in the Spark Library yet. When deciding what AI takes over, keep some steps with people because they teach the craft, even when the AI could do them.

Our reasoning, building on the idea that automation tends to leave people with the dullest and the hardest parts of the job and not much in between (Bainbridge, 1983).