Is 'I use ChatGPT' a good answer to 'how do you use AI at work'?

No. Naming a tool answers nothing. Interviewers want to hear the task, the judgment you applied, and the result you verified. The tool is the least interesting part of the answer — lead with the work and the outcome.

What if I only use AI a little?

That is fine — honesty beats inflation. Use the same three-part structure with a smaller example: one task you run through AI regularly, what you check, and what changed. A specific small truth outperforms a vague big claim every time.

Should I mention AI limitations in my answer?

Yes — briefly and confidently. Mentioning that you verify outputs, avoid confidential data, and know where the tool fails shows judgment, which is exactly what the question is testing. It also makes your answer sound like a real practitioner instead of a hype paragraph.

How do I answer this in a performance review vs an interview?

Same structure, different evidence. In a review, lead with results tied to your goals ('I cut weekly reporting time from 3 hours to 1 by automating the draft'). In an interview, lead with the task and the judgment you applied, and keep the outcome specific but brief.

What are the biggest mistakes in answering this question?

Three sink candidates: saying 'I let AI write it' (sounds like you outsource judgment), listing tools without tasks (sounds like a demo), and overclaiming with no specifics (falls apart under one follow-up question). All three are fixed by using the task-judgment-result structure.

How to Answer "How Do You Use AI at Work"

Somewhere between the second and third interview, it arrives: "So — how do you use AI at work?" It is now one of the most common questions for non-technical roles, and it is almost never a technology question. The interviewer is testing three things: whether you actually use AI on real tasks (not just in theory), whether you understand its limits (data privacy, hallucination, verification), and whether you sound like someone who would be safe and effective with it on their team. Most candidates fail not because they use AI too little, but because they answer with a tool name and nothing else.

This guide gives you a three-part structure that works in interviews, performance reviews, and even casual conversations with your manager. Then you get two complete example answers — one for an administrative role, one for HR — plus the mistakes that sink otherwise good candidates.

The framework: task, judgment, result

Every good answer follows the same three moves, in order:

Task. Name one concrete work task you use AI for. Not "everything" — one thing: meeting minutes, first drafts of emails, interview question prep, expense categorization, report formatting.
Judgment. Show the human decision you made around it. Which tool you chose and why, what you checked before using it (data safety, company policy), and what you verify in the output. This is the part that separates a practitioner from someone who just opened ChatGPT.
Result. Say what changed in the real world — time saved, fewer errors, faster turnaround — with one concrete figure if you have it.

Notice what this structure does: it puts the tool last and your judgment first. The interviewer learns that you treat AI as a draft engine you supervise, not as a substitute for thinking. That is the entire message they are listening for.

Example answer for an administrative role

"I use AI mainly for the three tasks that used to eat my week: meeting minutes, email drafting, and report formatting. For meetings, I take my raw notes and run them through our approved assistant with a prompt I wrote myself — it structures them into decisions, action items and open questions, and I always re-check the action items against the recording before sending. For emails, I draft the reply from my bullet points rather than the other way around, so the voice stays mine. The result is I cut the time on weekly reports from about three hours to under one, and I haven't missed a follow-up since I started using the action-item template. I keep personal data out of it entirely — nothing with client names or numbers goes into the tool."

Count the elements: specific tasks, a custom prompt, a verification step, a measurable result (3 hours to 1), and a data-safety line. That answer takes 45 seconds and tells the interviewer everything they need to know.

Example answer for an HR role

"In my current role I use AI most for job descriptions and interview prep. When a manager opens a role, I draft the JD from their bullet points and my notes about the team — AI structures it, I rewrite the tone and check every requirement against the actual duties before posting. For interviews, I generate a question set based on the competencies we agreed on, then remove anything that doesn't map to the role. I'm careful with candidate data — nothing personal goes into a public tool, and I use our approved workspace when we're handling applicant details. The measurable result: our time-to-post for a new role went from about four days to two, and hiring managers have commented the JDs are more consistent."

Again: task, judgment (JD checking, competency mapping, data handling), result with a number. The candidate sounds like someone who uses AI as a professional tool, not a toy.

How to tailor it to your own role

If you do not have a clean example yet, build one this week — it is the best career preparation you can do. Pick the task you repeat most, run it through AI once with a real prompt, and write down what you checked and what changed. Our prompt library has ready-made prompts for email, meetings, reports and interviews; the STAR answer generator turns your raw notes into the same structure an interviewer expects. If you want to see which AI skills your specific role needs first, the AI skills self-assessment scores you across five dimensions and shows your gaps — a useful way to find your first example task.

The six mistakes that sink candidates

"I let AI write it." Instantly reads as outsourcing judgment. Reframe: "I draft it, AI structures it, I verify and own it."
Listing tools instead of tasks. "I use ChatGPT, Claude, Gemini, Perplexity..." is a demo, not an answer. One task with depth beats five tools with none.
Overclaiming. "I've automated everything" collapses under one follow-up question ("what did you use for that?"). Keep it small and true.
Never mentioning limits. A candidate who says AI is flawless sounds naive. One line about verification or data safety adds more credibility than a paragraph of praise.
No result. "I use it for emails" leaves the interviewer to guess whether it worked. Add the before/after, even a rough one.
Ignoring company policy. Mentioning that you follow the approved tooling and data rules is a quiet signal that you are safe to hire.

Answering in a performance review

The review version of this question is usually phrased differently — "tell me about your work this quarter" — but the same structure applies, with one twist: lead with the result tied to your goals, then explain the method. For example: "One of my goals was faster reporting. I rebuilt our weekly report process around an AI-assisted draft, which took it from three hours to under one, and I documented the workflow so a colleague can run it on my days off." That one sentence does more for your rating than a paragraph listing tasks. The self-assessment guide shows you how to write the longer written version.

Answering when you're asked in a casual setting

Your manager asks over coffee. Do not launch into the full framework — give the one-line version: "Mostly for drafts and summaries — I give it my notes, it structures them, and I check everything before it goes out." If they are curious, they will ask for details, and then you can share the task-judgment-result version. Being ready for both registers is part of looking polished.

The one-sentence version you can memorize today

"I use AI as a supervised assistant on repetitive tasks — I give it my raw material, I set the rules, and I verify every output before it represents me." That sentence contains the task, the judgment, and the ownership in one breath, and it is a complete answer on its own if the interviewer is in a hurry. Add your specific example when they lean in.

Preparing your answer in 30 minutes

If an interview is coming up and you have not prepared, do this in half an hour. Minutes 1–10: list the five tasks you do most often, and mark which ones you already use AI for or could start using it for tomorrow. Minutes 11–20: pick the best one and write down your raw process — what you feed the tool, what you check, what used to happen without it. Minutes 21–30: turn it into the task-judgment-result structure in three sentences, and practice saying them out loud once or twice. That is genuinely all it takes: interviewers are not grading your AI sophistication, they are grading whether you have a real, repeatable story. If you want help shaping the story, the STAR answer generator walks you through the same raw-notes-to-structured-answer flow that interviewers expect.

Variants for other job families

The structure stays identical; only the example changes. Marketing: "I draft the first version of campaign copy from the brief and the audience notes, then I cut the AI-flavored phrases and check every claim against the product facts before it goes to review — it took our first drafts from half a day to about an hour." Sales: "I use AI to prep for calls — it pulls the account history into a one-page summary and suggests follow-up questions, and I always re-check the numbers against the CRM before I open my mouth." Finance: "I use it to structure reconciliations and draft the notes that explain variances — the calculations are mine, the narrative draft is faster, and I verify every figure." Customer support: "I use the knowledge base with AI to draft responses to repetitive tickets, then review each one for tone and accuracy — my handle time is down about 20 percent and satisfaction has not slipped." Pick the variant that matches the role, and keep the judgment element front and center.

Here is the reframe that makes the whole question easy: the interviewer is not looking for a power user, they are looking for a safe user — someone who gets the time savings without the data risk or the quality risk. Show them one task done well, one decision made consciously, and one result you can point to, and you will sound like the person they want on the team.