How to Answer 'How Have You Used AI at Work' in an Interview
Job interviews are catching up to how people actually work. Asking a candidate whether they've used ChatGPT used to be enough to check the AI box, but that question alone no longer tells employers much, according to a Veris Insights analysis published earlier this year. The firm found that surface-level questions like "What AI tools have you used?" measure awareness, not the kind of applied skill hiring teams actually want to see.
That shift matters for anyone preparing to answer how they've used AI at work in an interview, because the bar for a convincing answer has moved. Student use of generative AI in job searches has grown 2.6 times since 2023, and 62% of students say they feel pressure to use AI tools just to stay competitive, per the Veris Insights research. With more candidates name-dropping the same tools, employers say they need a better way to tell who can actually use them.
Some companies have already rebuilt parts of their interview process around this problem. Zapier has translated its expectation that all new hires demonstrate AI fluency into role-specific competency frameworks and interview questions, McKinsey has piloted case-interview formats where candidates use the firm's internal AI tool live, and Meta has adjusted technical interviews to see how candidates verify AI-assisted code, according to the same Veris Insights report. IBM, meanwhile, has said it plans to expand entry-level hiring while redesigning junior roles around AI-enabled workflows, shifting more early-career work toward judgment and customer interaction, per Veris, though the source doesn't describe IBM changing how it interviews candidates.
Not every employer has caught up. Many companies using AI still lack clarity internally about where or how it's actually being deployed, according to Keith Sonderling associate Justine Price, cited in a SHRM report from late last year. That gap between leading employers and everyone else means job seekers should expect real variation from one interview to the next.
What employers are actually testing for
Veris Insights describes the shift as one from tool familiarity to "applied capability," meaning interviewers care less about which software a candidate has opened and more about the judgment behind using it. The firm's research identifies four specific areas hiring teams are now probing.
- Prompting and tool use: whether a candidate can get meaningful output from AI and iterate when the first attempt falls short
- Critical thinking and validation: whether they question AI-generated content and can spot inaccuracies or gaps in it
- Workflow integration: whether AI is embedded in how they approach work or used only sporadically
- Communication and translation: whether they can explain how AI contributed to a result and collaborate around AI-assisted work
Candidates should expect follow-up questions built around these categories, such as "How do you validate the accuracy of AI-generated outputs?" or "What do you do when AI gives you a poor result?" according to Veris Insights. Because organizations range from early-stage informal assessment to fully redesigned, AI-enabled hiring, it's worth checking a job posting or company career page beforehand for any language about AI tools, workflows, or competencies. That gives a rough sense of how far along a given employer is before walking into the room.
How to answer "how have you used AI at work" using a five-part structure
A useful way to organize an answer is a five-part sequence: problem, process, verification, result, relevance. It mirrors what Veris Insights frames as the difference between outcome and process; interviewers are being told to focus on how a candidate approached a task, not just what came out the other end. Strong answers, per the same source, include specific examples, clear impact, and evidence of iteration. A line like "I use AI for brainstorming" doesn't hold up against that standard.
Here's what that structure sounds like in practice, using a marketing role as an example:
"In my last job, I had to turn three months of campaign data into a client-facing performance report on a tight deadline. I used an AI tool to generate a first draft, feeding it the raw metrics and a template from a previous quarter. Before sending anything out, I checked every figure against the original spreadsheet, caught two percentages the AI had pulled from the wrong column, and rewrote the executive summary because it read like generic marketing copy instead of something specific to this client. The report went out on time, and the client asked for the same format the next quarter. For this role, I'd take the same approach: let AI handle the repetitive first draft, then spend my time on the accuracy checks and client-specific judgment calls that actually need a person."
That answer works because it names a real problem, describes an actual process rather than a vague habit, shows a specific verification step (catching the mismatched percentages), states a concrete result, and closes by connecting the habit to the target job. It's also short enough to say without rambling. Drafting one or two examples like this ahead of time, built around a real task, gives a candidate something concrete to fall back on instead of improvising under pressure.
Tailoring your answer to your experience level
Candidates with extensive AI use in their current role should focus on showing that AI is woven into daily work rather than pulled out occasionally, since Veris notes interviewers are specifically checking whether AI is "embedded in how they approach their work, or used sporadically." A candidate who runs several tools through a regular workflow, adjusts prompts based on what didn't work last time, and can explain why they chose one tool over another for a given task is demonstrating exactly that.
Candidates with more modest exposure still have a workable path. Identifying one repetitive task, testing a tool against it, reviewing the output, and improving the result is a legitimate example, even if it's a single instance rather than a routine. Veris ties effective AI use to broader traits like problem-solving, adaptability, and a bias toward efficiency, all of which a limited example can still demonstrate if it's specific.
Candidates with little or no workplace AI experience, including those whose employers restrict AI tools, are better off describing how they would approach a task using AI rather than inventing a past example that didn't happen. Veris specifically recommends scenario-based interview questions because they "reveal how candidates think, not just what they know." A forward-looking answer that walks through a realistic task step by step can carry the same weight as a retrospective one, as long as it's specific about tools, checks, and reasoning rather than a vague statement of intent.
Talking about verification and data handling without overstepping
Interviewers evaluating AI skill are checking whether candidates can catch problems in AI-generated work, not just produce it. Veris's evaluation criteria include whether a candidate can identify inaccuracies or gaps in AI output, so mentioning a specific check, like the mismatched percentages caught in the sample answer above, tends to land better than a general claim of being careful. It gives the interviewer a concrete example of a review process rather than an assertion.
On data handling, it's reasonable to mention using employer-approved tools and avoiding entering confidential or client information into public AI platforms. That's a practical way to signal judgment without making a compliance claim that goes beyond what any candidate can actually verify in an interview setting.
the debate over human oversight in hiring is a separate issue from how candidates should describe their own AI-assisted work. Mike Bradshaw, vice president of talent at HR software provider Pinpoint, told SHRM last year that AI hiring tools can add real value but introduce risk if employers can't explain how a model reached a recommendation, adding that hiring decisions are judgments about people that should always sit with a human, according to the SHRM report. That principle applies to how employers use AI to evaluate applicants, not to how a candidate should talk about their own workflow. Separately, a study of 246 working-age adults published last year in the journal Media Psychology found broad skepticism toward employer claims that an algorithm alone can make an unbiased hiring decision, which is one more reason candidates should keep the focus on their own review process rather than making claims about AI's objectivity.
Before the next interview, check the job posting and any public statements from the employer about AI tools or workflows, then write out one specific example using the problem, process, verification, result, relevance structure and practice saying it aloud, aiming to keep it well under two minutes. If the posting doesn't mention AI expectations at all, that's worth asking about directly, since it's one of the clearest signs of where a given employer stands on the spectrum Veris describes, from informal to fully redesigned.
