Key Insights
- McKinsey’s AI interview started as a Lilli pilot. In late 2025 and early 2026, candidates in some U.S. final rounds worked a case with Lilli, McKinsey’s internal AI platform.
- Today’s version uses ChatGPT, and it counts. McKinsey told us in July 2026 that candidates use the free version of ChatGPT and the AI interview is just as important as every other McKinsey interview.
- Only some offices use it. It’s more common in Europe, it can replace one of your regular cases, and your recruiter tells you when to expect it.
- McKinsey is grading your judgment. The firm cares how well you question what the AI gives you, and your case fundamentals still carry the interview.
McKinsey now runs an AI-assisted case interview at some of its offices. You work a case with an AI tool for part of the interview, and the interviewer watches how you use it.
It began as the McKinsey Lilli interview, a pilot built around the firm’s internal AI platform. We first reported that pilot in December 2025. The format has changed since, and this guide covers what it looks like now and how to prepare.
What Is the McKinsey AI Interview?
The McKinsey AI interview is a case interview where you use an AI tool for part of the case. You work through a client problem, generate your own ideas, and review data, and at some point the interviewer brings AI into the conversation.
Simone Jacobi, McKinsey’s Director of Talent Assessment, walked us through the format in July 2026. Most interviews start without AI. The interviewer introduces the AI tool partway through, and the final part of the interview is AI-free again.
At the time of our conversation, candidates used the free version of ChatGPT.
From Lilli Pilot to AI Case Interview
Many candidates still call this the Lilli interview. Here’s how the format got from there to here.
- December 2025: MC reports McKinsey is piloting an AI interview in U.S. final rounds, mostly for Business Analyst candidates. Candidates work with Lilli, and the pilot is non-evaluative.
- January 2026: National press reports the Lilli pilot in final rounds for business school candidates.
- July 2026: McKinsey tells MC the AI interview is used at some offices, counts as much as any other interview, and gives candidates the free version of ChatGPT.
- September 2026: McKinsey’s Global Leader for Talent Attraction says the format is more likely in European offices and typically replaces one of the usual case interviews.
McKinsey hasn’t announced an end to the Lilli pilot. By July 2026, though, its own description of the format used ChatGPT. If you’re interviewing now, prepare for the AI case interview as McKinsey describes it today.
Where It Fits in the Interview Process
The AI interview isn’t standard at every office yet. Whether you get it depends on the office you applied to, and McKinsey says it’s more likely in European offices. Your recruiter will tell you when to expect it.
Where McKinsey uses it, the AI interview typically replaces one of your regular case interviews. It counts toward whether you advance, and Jacobi told us it’s just as important as every other McKinsey interview.
That’s a real change from the early Lilli pilot, which didn’t affect hiring decisions. Treat it like any other case in your round.
What Is Lilli?
Lilli is McKinsey’s internal generative AI platform, built on the firm’s own knowledge base of research and past client work. McKinsey consultants use it to research industries, structure problems, and draft early analysis.
Lilli was the tool in McKinsey’s first AI interview pilot. In the version McKinsey described in July 2026, candidates used ChatGPT. You don’t need access to Lilli to prepare, since the skills McKinsey grades are the same with any AI tool.
What the McKinsey AI Interview Looks Like
The interview feels like a normal McKinsey case for most of its length. You get a client problem, build a structure, work through data, and move toward a recommendation.
Partway through, the interviewer introduces the AI tool. You might use it to brainstorm drivers, test an assumption, or get a first draft of an analysis. Then you decide what to keep, what to question, and what to throw out.
The interviewer is watching that decision. Strong candidates are deliberate about when they use AI, what they use it for, and what they do with the output. The last part of the interview is AI-free, so you finish the case on your own thinking.
What McKinsey Is Really Testing
Jacobi told us prompt writing is a skill McKinsey can teach on the job. What the firm wants to see in the interview is your judgment.
"We are less interested in how much you use AI and more interested in how well you question what it gives you.
That line from Jacobi is the whole test in one sentence. In practice, McKinsey is looking for four things:
- Structured thinking: breaking down the problem before you ask the AI anything
- Judgment: spotting when an AI answer is generic, incomplete, or wrong
- Quantitative skill: checking the numbers yourself instead of trusting the tool
- Clear communication: explaining what you used, what you didn’t, and why
Jacobi was clear that core case fundamentals remain the foundation. AI makes them more visible, because you have to show your thinking about someone else’s output in real time.
How the McKinsey AI Interview Differs from a Traditional Case
In a traditional case interview, you generate all the analysis yourself, in a fairly straight line: structure, analyze, recommend.
In the AI interview, the middle of the case loops. You ask the tool for something, judge what comes back, refine, and fold the useful parts into your own answer. The extra skill is managing that loop without letting the tool set the direction of the case.
McKinsey AI Interview vs. Bain AI Interview: Key Differences Explained
McKinsey and Bain both now test how candidates work with AI, and the two formats look very different.
Bain’s AI interview, called the immersive case, gives you 40 minutes on your own with ChatGPT, Excel, and PowerPoint to work through a client’s files. Then you present for 10 minutes and defend your answer for 30. Our Bain AI interview guide has the full breakdown.
| Dimension | McKinsey AI Interview | Bain Immersive Case (AI Interview) |
|---|---|---|
| Status | Used at some offices, more often in Europe. Confirmed by McKinsey. | Reported by candidates in final rounds since fall 2026. |
| Does it count? | Yes. McKinsey says it’s as important as every other interview. | Bain hasn’t said how it’s weighted. Candidates report it as one of three final-round interviews. |
| Tool | Free version of ChatGPT (early pilots used Lilli). | ChatGPT, plus Excel and PowerPoint. |
| Where it appears | Typically replaces one of your regular case interviews. | Final rounds, alongside a regular case and a behavioral interview. |
| What candidates do | Work a live case with the interviewer, using AI for part of it. | Work on their own for 40 minutes with client files, then present for 10 minutes and discuss for 30. |
| What the interviewer sees | How you use and question the AI, in real time. | What you built and how well you defend it. |
| Candidate takeaway | Use AI on purpose and question everything it gives you. | Build a clear recommendation fast, check every number, and be ready to defend it for 30 minutes. |
The practical difference: at McKinsey, your process is on display while it happens. At Bain, your process shows up in what you built and how well you explain it. See every AI format at McKinsey, Bain, and BCG in our AI case interview tracker.
How to Prepare for the McKinsey AI Interview
Start with your fundamentals. The AI interview is still a McKinsey case, and the AI portion won’t rescue a weak structure or shaky math.
Then practice cases with the free version of ChatGPT open, since that’s what candidates have used. A few habits to build:
- Structure first, then ask. Know what you need from the tool before you type anything.
- Question every output. For each answer, ask yourself what’s useful, what’s missing, and what you’d change before a client saw it.
- Check the numbers yourself. ChatGPT gets math wrong. Recompute anything that will drive your recommendation.
- Talk through your choices out loud. Say why you asked what you asked and why you kept or dropped what came back.
- Practice finishing without AI. The last part of the interview is AI-free, so you need to land the recommendation on your own.
Common Mistakes to Avoid
The biggest mistake is letting the tool drive. When the AI’s output becomes your structure, the interviewer can tell right away that you aren’t in control of the case.
The second is accepting polished answers at face value. AI output often sounds confident and still misses the point. Pause, question the assumptions, and test the conclusion before you build on it.
The third is never landing the plane. Candidates explore ideas with the tool and run out of time before stating a clear recommendation. Every McKinsey case still ends with your answer to the client’s problem.
Why McKinsey Added AI to Its Interviews
McKinsey teams already use AI to get to a first draft faster, which leaves more time for judgment and creativity. Jacobi told us the AI interview is meant to reflect that way of working.
McKinsey has also said it may build more AI-related elements into its assessment over time. It’s already piloting an AI client avatar in some business analyst interviews, its newest format.
Black Belt
Practice the McKinsey AI interview before it counts
Black Belt pairs you with an MBB coach who can run AI-assisted cases with you and push back the way a McKinsey interviewer will. You also get the AI Interview Prompt Playbook and an AI case coach to practice with between sessions.
See how Black Belt worksThe Bottom Line
McKinsey’s AI interview started as an unscored Lilli pilot. It’s now a case interview that counts, at the offices that use it.
Ask your recruiter whether your round includes it. Either way, practice a few cases with ChatGPT open, so the tool feels normal and your judgment is what the interviewer remembers.
McKinsey AI Interview (Lilli): Frequently Asked Questions
The McKinsey Lilli interview was McKinsey’s first AI interview pilot, where candidates in some U.S. final rounds worked a case with Lilli, the firm’s internal AI platform. It has since become McKinsey’s AI case interview, where candidates use the free version of ChatGPT for part of a case.
McKinsey hasn’t announced an end to the Lilli pilot. In July 2026, McKinsey told us candidates in its AI case interview use the free version of ChatGPT.
Yes. The early Lilli pilot was non-evaluative, but McKinsey told us in July 2026 that the AI interview is just as important as every other McKinsey interview. Where it’s used, it typically replaces one of the regular case interviews and counts toward whether you advance.
Only some offices use the AI interview, and McKinsey says it’s more likely in European offices. Your recruiter will tell you if your round includes it.
The McKinsey AI interview is a case interview where the interviewer introduces an AI tool partway through. Candidates use it for part of the analysis, judge what it produces, and finish the case without AI.
Practice McKinsey-style cases with the free version of ChatGPT open. Structure the problem before you prompt, question every output, check the numbers yourself, explain your choices out loud, and practice finishing the case without AI.
McKinsey tests structured thinking, judgment, quantitative skill, and clear communication. McKinsey doesn’t grade prompt writing; it cares more about how well you question what the AI gives you than how much you use it.