Hiring your first AI lead

The most expensive hiring mistake I watch scaleups make is hiring a researcher when they needed an operator.

A research background signals depth on model behaviour. It signals nothing about whether someone can sequence a roadmap, say no to a founder, negotiate with a data protection officer, or ship something small on time.

For a first AI hire in a company under 300 people, weight judgement under ambiguity above everything. Can they tell you which of your use cases should not be built? Can they explain a failure mode to a non-technical board? Have they ever turned something off?

Ask candidates for a system they retired and why. The answers are far more revealing than the ones about systems they launched.