LLMs in interviews
First posted on LinkedIn.
We’ve been encouraging candidates to use LLMs during our face-to-face interviews. One thing is becoming clear: LLMs must not become a crutch.
In interviews — and increasingly in day-to-day work — the model is missing context. It doesn’t have the internal domain knowledge, does not see the business constraints, historical decisions, or second-order trade-offs. As a result, the output is often incomplete, filled with noise, or directionally wrong.
The real work still sits with the human:
- Asking good questions to surface missing context.
- Structuring and pruning inputs before prompting.
- Discerning what in the output is signal versus noise.
- Critiquing the response — what’s right, what’s wrong, and what can be improved.
If you use an LLM during an interview, don’t just read its output.
Share your screen and review the response together with the interviewer — what’s useful, what’s wrong, and how you would improve it.
Remember: interviewers are evaluating your thinking and how you use all the tools to augment and amplify your own abilities.
One final note: LLMs are most useful for case studies and unfamiliar scenarios.
When talking about past work, relying on an LLM can get in the way. Your own experiences already contain the context — the decisions you made, the trade-offs you navigated, and the outcomes you owned.