Building Xcellerant's seventeen-person AI product team meant interviewing a lot of candidates whose resumes said "product manager" and a few whose resumes didn't (business analysts, engineers, one person moving over from customer support). Experience predicted less than I expected. What predicted more was something closer to temperament, and it took me a while to learn how to actually test for it in an hour-long interview.
The signal that mattered more than the resume
The single best predictor I've found is whether someone, handed an ambiguous problem on the spot, reaches for Comprehend before they reach for a solution. Most candidates, including experienced ones, will start proposing fixes within the first two minutes of a case question. The ones worth hiring ask a clarifying question first, then another, visibly resisting the pull toward sounding decisive before they actually understand the situation.
I'm not hiring for confidence. Confidence is cheap and most candidates have learned to perform it. I'm hiring for the much rarer willingness to look uncertain in front of a stranger, on purpose, because the alternative is being wrong faster.
What I actually ask
- A real, messy situation from my own history, not a hypothetical. I want to see how they handle genuine ambiguity, not a rehearsed framework.
- "What would change your mind?" after they give an answer. Candidates who can't name a specific piece of evidence that would reverse their own recommendation usually haven't actually reasoned their way there. They've pattern-matched to something that sounds right.
- How they talk about a past disagreement with an engineer or designer. Not whether they won it, but whether they can describe the other person's position as something a reasonable person could hold, rather than an obstacle.
Where titles actively mislead
The strongest hire from that whole process had spent three years in customer support, not product. She'd never written a PRD. She had, however, spent three years listening to the exact moment a product failed someone, in their own words, dozens of times a week, which turned out to be a better training ground for Comprehend and Identify than two years of writing specs from a comfortable distance. I'd have screened her resume out if I'd been hiring for the title instead of the skill underneath it.
Of the people I hired and coached into the 17-person team, the strongest performers a year in weren't the ones who arrived with the most PM experience. They were the ones who arrived asking the most uncomfortable questions in round one.
The thing I actively coach out of new hires
The opposite failure mode is just as common: candidates who are so comfortable with ambiguity that they never converge. The job isn't "ask good questions forever." It's ask good questions until you can defend a specific, falsifiable recommendation out loud, on a deadline. I'm hiring for both halves: the discomfort to slow down, and the discipline to still land the plane.