Mark Parfenov · Updated

Guide · AI and your work

Explain your own work when AI helped build the project

Interviewers now ask how you used AI. Show what you decided, checked and owned, tell it as a clear story, and answer the follow-up questions without underselling or overselling yourself.

6-minute read · Free · Last part unlocks with your email · By Mark Parfenov · Published 26 October 2026

Not sure how to talk about your AI-assisted work? Message me: TelegramWhatsApp

“Did you use AI for this?” is now a normal interview question. Most people answer it badly in one of two ways: they hide the tools and get caught on a detail, or they credit the tools with everything and disappear from their own story. The good answer is in between: the tool did some typing, and you made the decisions.

01Why interviewers ask

AI tools are now ordinary at work. In the Stack Overflow 2025 Developer Survey, 84% of developers said they use or plan to use AI tools, and more of them distrusted the accuracy of the output than trusted it. Interviewers know both things.

So the question is rarely “did you use AI?”. It’s “can you still think, decide and check?”. Your answer should make those three things easy to see.

02Separate what the tool did from what you did

Before any interview, take one project and split it into two columns. The left column can be long; your story lives in the right one.

What the tool didWhat you can say about your part
Generated the boilerplate for the API endpoints“I designed the API: which endpoints, what they return, how errors work.”
Suggested a caching approach“I compared it with two options and chose Redis because of our read pattern.”
Drafted unit tests“I decided what had to be tested and added the edge cases it missed.”
Proposed a fix for the memory leak“I confirmed the cause in the profiler before accepting the fix.”
Wrote the first version of the docs“I rewrote them for the support team and checked every step.”

If the right column is empty for a project, choose a different project for the interview.

03Four things only you can claim

  1. The problem

    You noticed it, framed it and decided it was worth solving.

  2. The decisions

    Architecture, trade-offs, what to build and what not to.

  3. The checking

    Tests, reviews, real data, the bug you caught.

  4. The responsibility

    You shipped it, supported it and answered for the result.

These are what employers pay for. Mention at least three of them in every AI-assisted story.

04Tell it as a PAR story

Use Problem, Action, Result, and put the tools inside the Action, as one detail among others.

Weak“I built an internal dashboard with ChatGPT and Copilot. AI wrote most of the code, so it was quite fast.”
Strong“Support was answering the same refund questions every day. I designed a dashboard that shows each order’s refund status. I used Copilot for the React components, but I designed the data model, wrote the SQL myself and tested it against last month’s cases. One AI-generated query double-counted partial refunds; I caught it in testing. Support tickets about refunds fell by about a third.”

The strong version mentions AI twice, honestly, and still makes it obvious who built the system. It also contains one moment of judgment: the bug the tool introduced and you caught.

05Prepare for the follow-up questions

Expect at least one of these, and prepare a two-sentence answer for each:

  • “How did you check the output?” Name the method: tests, real data, a profiler, a review.
  • “What did the tool get wrong?” Have one specific example. It shows you were paying attention.
  • “What would you have done without it?” The same design, more slowly. Say which parts.
  • “Can you explain this part of the code?” Only include code in your story that you can explain line by line.

06Stay honest and inside the rules

  • Never claim code you can’t explain. A live follow-up will find it.
  • Don’t hide the tools either. If it comes out later, it looks worse than the use itself.
  • Mention data rules if they applied: “We don’t put customer data into external tools, so I used synthetic data.”
  • Follow the interview’s own rules. Some coding rounds allow AI and some don’t; ask if it isn’t clear.

07Phrases for AI-assisted work

SituationPhrases
Saying how you used the tool“I used Copilot for the boilerplate; the design and the data model were mine.” · “I used AI to draft the tests, then decided what was actually worth testing.”
Showing your decisions“I chose X over Y because…” · “I rejected the first suggestion because it didn’t handle…”
Showing how you checked“I checked it against real data from last month.” · “I reviewed every line before it went to production.” · “I confirmed it in the profiler.”
What the tool got wrong“It suggested a query that double-counted refunds. I caught it because the totals didn’t match the finance report.”
The “without AI” question“Without the tool, the components would have taken longer, but the approach would be the same.” · “I can walk you through how it works.”
Data and company rules“We don’t put customer data into external tools, so I used synthetic data.” · “I followed our team’s AI policy.”
Instead ofTry
“AI did most of it.”“AI wrote the first draft of the components; I designed and tested the system.”
“I just used ChatGPT.”“I used ChatGPT to compare three approaches, then chose…”
“I don’t use AI at all.” (when you do)“I use it for boilerplate and first drafts; I review everything.”
“It works, I think.”“I tested it with these cases, and here’s what I checked.”

08Example: a project answer in an interview

An invented example. Notice where the tools appear: inside the action, next to the decisions and the checking.

Internal refunds dashboard

Interviewer: Tell me about a recent project you’re proud of.

Candidate: Our support team spent hours every week answering the same refund questions, because nobody could see a refund’s status in one place. I proposed and built an internal dashboard to fix that.

Candidate: I designed the data model and wrote the SQL that joins orders, payments and refunds. For the React components I used Copilot, which saved me a few days. I also used ChatGPT to compare two charting libraries.

Candidate: Before launch, I tested every query against last month’s real cases. One AI-generated query double-counted partial refunds; I caught it because the totals didn’t match the finance report, and I rewrote it.

Candidate: After launch, refund tickets fell by about a third, and the support lead now uses the dashboard in her weekly report.

Interviewer: What would you have done without AI?

Candidate: The same design, just slower on the front end. The decisions and the testing would be exactly the same. I’m happy to walk you through any part of the code.

09Before your interview

  • I’ve split my main project into “the tool” and “me”.
  • My story names the problem, my decisions and how I checked.
  • I have one example of the tool getting something wrong.
  • I can explain every piece of code I mention.
  • I have a two-sentence answer to “what would you do without it?”.
  • I know the data rules I followed.
AI and your work

Want to practise telling your project stories?

I help people in tech turn their projects, AI-assisted or not, into clear interview stories in English, and practise the follow-up questions until the answers come easily.

Book a free consultation

Questions first? Message me: TelegramWhatsApp

Free guide

Enter your name and email to download the guide. I’ll also email you a copy.

I use your details to send you the guide, and I may follow up once to offer a free consultation. See the privacy policy.

Choose a time