To explain a technical project in an interview with AI, write a plain fact sheet about the project, then use a chatbot to build three versions of it (a 30-second headline, a two-minute walkthrough and a deep-dive outline), to test the explanation on a non-technical listener and to grill you on your decisions and trade-offs. It works for engineers describing a system and for analysts, marketers and operations people describing a dashboard, an automation or a rollout. AI is good at restructuring and simplifying. Technical accuracy and confidentiality are on you: check every detail it adds, and keep internal names, security details and client data out of the chat.
Below: the three versions, how the explanation changes by listener, a seven-step workflow, five prompts, one project explained two ways, and the accuracy checks. For the long-form round, see what a project deep dive interview is. For AI-assisted prep in general, see how to use AI for interview prep.
Three versions of the same project
| Version | Length | When you'll need it | What it must include |
|---|---|---|---|
| Headline | About 30 seconds | Recruiter screens, "tell me about yourself", panel introductions | The problem, your role, the result |
| Walkthrough | About 2 minutes | "Tell me about a project you're proud of", hiring manager rounds | Context, the hardest part, one decision and why, the result |
| Deep dive | 10 minutes or more, with interruptions | Project deep dives and technical rounds | How it works, alternatives you rejected, what failed, what you'd change |
Most people prepare only the walkthrough, then ramble in a recruiter screen and run dry in a deep dive. Building all three from one fact sheet takes about an hour with AI.
Who's listening changes the explanation
| Listener | Cares about | Cut | Add |
|---|---|---|---|
| Engineer or technical peer | How it works and why you chose it | Basic definitions | Trade-offs, numbers, failure modes |
| Hiring manager | Your judgment and effect on the team | Detail they didn't ask for | The decision you made and what it changed |
| Recruiter or HR | Scope, your role, the outcome | Acronyms and tool names | A plain before and after |
| Senior leader | Effect on the business | Mechanism | Time, risk or cost in plain terms |
Technical projects aren't only for engineers. A marketing analyst who built an attribution dashboard, an operations lead who rolled out handheld scanners in a warehouse and a finance analyst who automated a month-end reconciliation all face the same problem: explaining it to someone who doesn't share their vocabulary. Our guide to "tell me about a time you explained something complex" covers the behavioral version of that skill.
Workflow: fact sheet to three versions
- Write a fact sheet without AI: the problem, the constraints, your part and others', two or three decisions with the alternatives you considered, what went wrong, the result and how you measured it, what you'd change.
- Sanitize it. Replace internal system names with generic ones ("our order service", "the warehouse system"); remove client names, security details and anything unreleased.
- Build the three versions with prompt 2.
- Test the walkthrough on a non-technical listener (prompt 3) and fix every word it didn't follow.
- Grill your decisions (prompt 4). This is where deep dives are won or lost.
- Explain it back from memory, then compare with the fact sheet (prompt 5).
- Rehearse with interruptions (see the last section).
Five prompts to copy
1. Fact-sheet interview
Interview me about a project I want to discuss in interviews. Ask one question at a time: the problem, the constraints, what I did versus others, two decisions and the alternatives I considered, what went wrong, the result and how it was measured, and what I'd change. Don't fill in any technical detail I don't give you. At the end, write a fact sheet using only my answers.
Why it works: the "what I did versus others" question settles the I-versus-we issue before you start drafting.
2. Three versions
Using only this fact sheet, write: (a) a 30-second headline of about 75 words; (b) a 2-minute walkthrough outline in 6 bullets; (c) a deep-dive outline listing the questions an expert interviewer would likely ask at each point. Mark anything you inferred rather than read in the fact sheet. Fact sheet: [paste]
Why it works: the inferred-detail markers show you exactly where the AI went beyond your facts.
3. The non-technical listener
Act as a smart listener with no technical background, like a recruiter. I'll paste my 2-minute explanation. List every word or step you didn't understand, then ask me two questions about why the project mattered. Don't rewrite it for me. Explanation: [paste]
Why it works: you fix the explanation in your own words instead of adopting the AI's.
4. Decision grill
Act as a senior [engineer / analyst / operations manager] interviewing me about this project. Take one decision at a time: ask why I chose it, what I rejected and why, what it cost, and what would make me change it. Push back once on each answer. Fact sheet: [paste]
Why it works: expert interviewers care more about the reasons behind a choice than the choice itself.
5. Explain-back check
Here's my fact sheet and a transcript of me explaining the project from memory. List anything in the transcript that contradicts the fact sheet, any claim stronger than the fact sheet supports, and any important point I left out. Fact sheet: [paste]. Transcript: [paste]
Why it works: it catches the drift that creeps in when you retell a project many times.
Example: one project, two listeners
Weak version for a recruiter (example we wrote): "I migrated our nightly batch jobs from cron on a single VM to an orchestrated DAG setup with retries and idempotent tasks, added data-quality assertions on the staging tables, and moved alerting from email to our paging tool. That cut our MTTR significantly and eliminated most of the partial-load incidents we were seeing from upstream schema drift."
What prompt 3 found: seven terms the listener didn't follow (cron, VM, DAG, idempotent, staging tables, MTTR, schema drift) and no answer to why it mattered.
Improved version for a non-technical listener (example we wrote): "Every night our systems copy the day's orders into the reports the sales team reads each morning. That process ran on one machine and failed quietly a few times a month, so people sometimes made decisions on half-finished numbers. I rebuilt it so each step can safely re-run if it fails, added checks that stop bad data before it reaches the reports, and made failures alert the on-call engineer instead of sending an email nobody read. Failures became rare, and when one happens we now fix it before the sales team starts work."
For a technical interviewer, keep the original vocabulary and add the reasons: why an orchestrator rather than patching the scripts, how you made tasks safe to re-run, and what you'd measure next. Same project, same facts, different emphasis.
Checking for accuracy and confidentiality
- Added technology. Chatbots sometimes add tools or patterns you didn't use. Delete anything that isn't on your fact sheet; an expert will ask about it.
- Slightly wrong explanations. A technique described almost correctly is worse than one left out. Check every technical sentence against what you know.
- Inflated scale. "A lot of traffic" must not become "millions of users".
- Oversimplified to the point of untrue. The plain version still has to be accurate. "Made it faster" is wrong if what you did was make it reliable.
- Confidentiality. Internal system names, security setup, client data and unreleased products stay out of the chat and out of the interview. Check your employer's AI-tool policy and any confidentiality agreement you signed.
Rehearse with interruptions
Real interviewers interrupt. Ask a chatbot to stop you after every two or three sentences with a question, and practice picking up where you left off. Practice describing your architecture or workflow in words ("three parts: where data comes in, where it's checked, where it's reported"), and if the interview includes a whiteboard or diagram tool, practice sketching it in under three minutes.
Engineers should also link the project to their history; our guide to walking through a software engineer's résumé shows where it fits, and "tell me about a time you solved a difficult problem" is often the question that opens the door to it. Then say the walkthrough under pressure in an AI mock interview built from your résumé and the job description. In live interviews, follow your employer's rules.
Where CoPilot Interview fits
CoPilot Interview is an AI interview assistant whose AI mock interviews read questions aloud, draw on your résumé and the job description, and score each spoken answer out of 10, which makes them a good check on whether your two-minute walkthrough is clear. The free plan gives 15 minutes a day of live help in sessions of up to 15 minutes, with AI mock interviews included. Paid plans are Standard at $24.99 a month and Pro at $39.99 a month. It's available as a desktop app for Windows 10 and 11 and macOS 12 or later, a free iPhone and iPad companion on the App Store (iOS 16 or later) and an Android companion in beta.
For a quick try, the free practice demo includes technical and behavioral tracks, scores typed answers and asks a follow-up, with no signup. In real interviews, follow your employer's rules.
Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly.
Frequently asked questions
How do I explain a technical project to a non-technical interviewer?
Lead with the problem in everyday terms, say what you did in one or two plain sentences, and finish with what changed for the people affected. Cut acronyms and tool names unless asked, and check the simple version is still accurate.
How long should I talk about a project in an interview?
About 30 seconds for a headline, about two minutes for a walkthrough, and ten minutes or more in a project deep dive, where the interviewer will interrupt with questions.
Can I use ChatGPT to prepare my project explanation?
Yes, to restructure, simplify and question it. Write the facts yourself first, remove anything confidential, and delete any detail the AI adds that you can't explain.
What if I'm not an engineer but my project was technical?
The same method works: a dashboard, an automation, a system rollout or an analysis. Build the three versions and practice switching between a specialist and a general listener.
How do I talk about a team project without taking all the credit?
Say what the team did with "we" and what you did with "I", and name the decisions you made. Interviewers want to know your part, and being precise about it reads as honest.
Rehearse it out loud before the real one
The AI mock interview builds questions from your resume and the job description, reads them aloud and scores each spoken answer out of 10. Included on the free plan.
Start a mock interview