Tell me about a technical decision you got wrong on an AI project. How did you find out, and what did you do?
What they are really testing: Calibration. In a field this young everyone has been wrong recently, so a candidate with no story is either not building or not reflecting, and the quality of the recovery matters more than the mistake.
A real interview question
Tell me about a technical decision you got wrong on an AI project. How did you find out, and what did you do?
What most people say
drag me
“I once picked a model that turned out to be too slow, we switched to a faster one once we noticed, no big deal in the end.”
The safest possible mistake, discovered passively, fixed trivially, with no reflection. It answers the question grammatically and refuses it substantively, which interviewers read as either inexperience or defensiveness.
The follow-ups they ask next
How did you handle having publicly argued for the losing option?
Calling it yourself, early and with data, converts a credibility loss into a credibility gain. The expensive version is defending it two quarters too long.
What did you keep from the failed approach?
The eval harness and the curated data outlived the decision. Sunk work is often partially salvageable, and naming what transfers softens the write-off honestly.
What the interviewer is listening for
- Owns a decision they argued for, with the reasoning that seemed right
- Falsified by measurement and named the assumption that broke
- Extracted a rule about trend-modelling and iteration speed as cost
What sinks the answer
- No real mistake available, or a trivial one dressed up
- The discovery was passive and the reflection absent
- Blames the field moving fast rather than the static assumption
If you genuinely do not know
Say this instead of freezing. Reasoning out loud from what you do know beats silence every single time, and a good interviewer is listening for exactly that.
“Shape: [the decision and why it looked right], [the signal that falsified it and how fast], [unwinding it yourself, including the retro], [the rule: model the trend, price iteration speed, date-stamp the assumptions].”
Keep going with behavioural
Junior
This field changes monthly. Tell me about a time something you had built became outdated fast, and how you handled it.
Mid
Tell me about a time an AI feature you shipped behaved badly in production. What happened and what did you change?
Mid
Describe a time you pushed back on using AI for something. How did you make the case, and what happened?
Mid
Describe a time you were pressured to ship an AI feature before you thought it was ready. What did you do?
Senior
Tell me about a time your eval data said one thing and an important stakeholder insisted the opposite. How did you resolve it?
Senior
Tell me about an AI feature that worked technically but users did not adopt or trust. What did you learn?
Knowing the answer is not the same as recalling it under pressure
Sign in to send the questions you fumble to spaced recall, so they come back right before you would forget them, and learn the concepts behind them with hands-on labs.
Start free