Every quarter brings a new AI paradigm the company is urged to adopt. As the senior AI voice, how do you decide what the organisation adopts, watches, or ignores?
What they are really testing: Whether they have a repeatable filter that spends the org innovation budget well, and whether they can hold the line against both hype and dismissiveness with visible, criteria-based decisions people trust.
A real interview question
Every quarter brings a new AI paradigm the company is urged to adopt. As the senior AI voice, how do you decide what the organisation adopts, watches, or ignores?
What most people say
drag me
“I stay on top of developments and evaluate promising technologies, adopting what fits our stack after proof-of-concepts, balancing innovation and stability.”
A paragraph of balance-speak with no mechanism. Who triages, against what criteria, how a spike is bounded, what watch concretely means, none of it exists, and a leadership answer without a mechanism is a mood.
The follow-ups they ask next
The CEO read about agent swarms and wants a company initiative. Walk me through the conversation.
Run it through the visible mechanism, not a hallway no: here is the problem list, here is what a 2-week spike would test against our evals, here are the criteria. Leadership pushes respond to a legible process far better than to an expert frown.
How do you keep the triage rubric itself from ossifying?
The quarterly watch review doubles as a rubric review: check the ignores that turned out wrong, ask what criterion would have caught them, and adjust. The filter is versioned like any other artifact.
What the interviewer is listening for
- Inverts the direction: business problems rank technologies
- Cheap bounded mechanism: triage page, timeboxed spikes, pre-written criteria
- Ignore-with-reasons and watch-with-triggers, publicly legible
What sinks the answer
- Pure enthusiasm or pure scepticism with no repeatable mechanism
- Proof-of-concepts on vendor demo paths rather than own evals
- Decisions private and personal, so trust in them depends on trust in one person
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.
“Mechanism over mood: [a living list of expensive business problems ranks every new paradigm], [one-page triage then a timeboxed spike against our own evals with criteria written first], sorted into [adopt with owner and kill criterion, watch with named triggers, ignore with written revisitable reasons], all [published internally so the process, not my taste, carries the authority].”
Keep going with strategy
Senior
The CTO asks whether you should build your AI capability on frontier APIs or self-host open-weight models. How do you frame the decision?
Principal
You have 8 product teams all building AI features independently. Do you build a central AI platform team, and if so, what does it own?
Principal
The board asks: we spent 2 million dollars on AI this year, what did we get? How do you make AI investment measurable, before and after the spend?
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