Preparing for interviews with AI

Generic lists of interview questions are close to useless, because interviewers ask about the role in front of them. Paste the actual job description and ask for the fifteen questions a hiring manager for this specific role would most likely ask.

You will typically get ten predictable ones and a handful you had not considered. Those last few are the entire value of the exercise — they are the questions you would otherwise have encountered cold.

Push it one step further by asking which requirement in the posting is hardest to verify in an interview. That is almost always where the deepest questioning lands, because it is the thing the hiring manager is most worried about getting wrong.

Rather than preparing an answer per question, prepare six to eight stories from your experience and know each one well. Most behavioural questions are variations on a small set of themes, and the same story often answers several of them from different angles.

For each story keep the situation, what you specifically did, the outcome, and what you would do differently. That last part is what distinguishes a reflective candidate from a rehearsed one, and interviewers consistently probe for it.

Write each one down once, in rough notes rather than prose. The act of writing fixes the details — the numbers, the timeline, who else was involved — so you are not reconstructing them under pressure while also trying to sound coherent.

• Something you shipped end to end
• Something that failed, and what you changed afterwards
• A disagreement with a colleague and how it resolved
• A time you worked with incomplete information
• Something you learned quickly under pressure
• A decision you made that turned out wrong

The most underused move is adversarial. Give it one of your stories and ask what a sceptical interviewer would push on, or where the account sounds like you are claiming more credit than you had.

The follow-ups it produces are usually the exact ones that catch people out: what would have happened if you had done nothing, what did the rest of the team think, how did you know the improvement came from your change rather than something else.

Answering those in advance is worth more than polishing the original story. An interviewer is rarely persuaded by the first answer; they are persuaded by whether it holds up when they press on it.

Reading a memorised answer is obvious to anyone who interviews regularly. It also removes the thing that actually persuades people, which is watching you reason through something in real time.

Say your answers aloud instead, from bullet points, and let the wording differ every time. If an answer runs past two minutes, it is too long — practise finding the shorter version.

Record one attempt and listen back once. It is uncomfortable and it is the fastest way to notice the filler, the tangents and the point where you answered a different question from the one you were asked.

Ask for a briefing on what the company does, who its customers are and what has changed recently, then verify the important claims yourself. Models get details about specific companies wrong with some regularity, and repeating a wrong fact in an interview is worse than not raising it.

What you are looking for is enough context to ask a good question and to explain why this role, specifically, rather than any role.

Treat anything with a date attached — a funding round, a launch, a leadership change — as needing a source before you use it. Those are exactly the facts a model is most likely to state confidently and get wrong, and exactly the ones an interviewer will notice.

The questions you ask are assessed, whether or not the interviewer says so. Vague ones about culture signal that you have not thought about the work; specific ones about how the team operates signal that you have.

The strongest questions are the ones only someone who read the posting carefully could ask — about a responsibility that seemed unusual, or about how two listed priorities are balanced when they conflict.

• What does the first ninety days look like for this role?
• How does work get prioritised when everything is urgent?
• What is currently the most frustrating part of this codebase or process?
• How do you know when someone in this role is doing well?
• What would make you regret this hire in six months?

Spend ten minutes writing down what was actually asked while it is fresh. Over a search of several interviews this becomes the most valuable preparation material you have, because it is drawn from real rooms rather than from a generated list.

Note anything you answered badly and write the better version now. The same question recurs across companies far more often than people expect, and the second time should not go the same way.

If a specific gap came up twice, treat that as a signal rather than bad luck. It usually means either the CV is pointing you at roles slightly beyond your evidence, or there is one concrete thing worth learning before the next round of applications.