Tell me about a time you disagreed with a senior leader about shipping an AI feature.
- 1 min — situation: the feature, the date, the metric gap
- 2 min — how and where you raised it, the leader's objection
- 2 min — the alternative you proposed and built
- 1 min — result with numbers, and the regret
The interviewer is testing whether you disagree with data, in the right forum, and then commit. Strong answers show the leader's goal was legitimate, your concern was measured not asserted, and you offered a path that met their goal with less risk.
Use a story where the disagreement was about readiness — hallucination rate, a missing eval gate, an unbounded blast radius — and where you proposed a staged rollout rather than a veto.
At Staff level the disagreement is about a decision with org-wide consequences, and the resolution changes a process, not just one launch. Show you understood the leader's pressure (a customer commitment, a board date), and that the outcome was a launch, not a delay won by attrition.
What the answer sounds like at each level
Disagreed with a lead on a component they owned, used data, and accepted the outcome.
Disagreed with a product leader on a launch, offered a staged alternative with criteria, and delivered it.
Changed how launch readiness is decided across teams — per-slice criteria and staged rollouts became the default — while still meeting the business date.
Follow-ups the interviewer will push
1What if the VP had overruled you?
I'd commit and make the risk visible: write down the concern and the trigger that would roll back, put the dashboard in front of the VP weekly, and own the response if it fired. Disagree-and-commit means committing for real.
2How did you keep the relationship intact?
I framed it as protecting their launch, brought options not a veto, and credited the VP's date as the forcing function in the retrospective — which it genuinely was.
Red flags interviewers listen for
- Leader portrayed as foolish; no acknowledgement of their legitimate goal.
- The 'win' is a delay with no launch plan.
- No numbers — 'quality wasn't good enough'.
How it's scored
| Dimension | What strong looks like |
|---|---|
| Data | Per-slice metric with n, not an aggregate. |
| Influence | Private forum first, options not veto, persuaded the dissenter on their terms. |
| Ownership | Clear about what they did vs the team, and a real regret. |
Shipping clause extraction to 10% instead of 100%
Illustrative exampleOur VP of Product wanted to launch AI clause extraction for contract review to all 1,200 customer accounts on a fixed date tied to a sales event. On our 500-contract golden set, extraction was 91% correct overall but only 74% on indemnity and limitation-of-liability clauses — the two that matter most to lawyers.
I was the tech lead for a team of six. My job was to give an honest launch recommendation and, if I disagreed, a plan the VP could accept.
I didn't argue in the launch meeting; I asked for 30 minutes the next day and brought a one-page breakdown by clause type with ten real failure examples. The VP's objection was that 91% was better than the competitor's demo. I agreed it was, and proposed launching on the date to 10% of accounts with low-confidence indemnity extractions flagged for human review, ramping to 100% when that slice reached 88%. I wrote the rollout plan; the team built the confidence flag in eight days and a per-clause dashboard. I also committed publicly that if we missed the ramp criteria I'd present the reasons myself.
We launched on the date to 120 accounts. Indemnity accuracy reached 89% after three weeks of error analysis on flagged cases, and we went to 100% in week five. In that window we caught 31 extractions that would have been wrong in front of customers. What I'd do differently: I raised the concern four weeks before launch; raising it at spec time would have made it a plan input, not a late objection.