McKinsey AI Interview: What It Is and How to Prepare
McKinsey is adding an AI interview component in 2026. Here's exactly what it tests, how it works, and how to prepare.
McKinsey is changing its interview process in 2026, and most candidates preparing right now have no idea.
The firm is piloting an AI-driven interview component, powered by its internal tool Lilli, as part of final-round recruiting. This guide explains exactly what that component is, what it actually tests, how it differs from a traditional case, and what you need to do in the next six to ten weeks to be ready for it. Read this once and you will know more about this format than 90% of the candidates sitting across from it.
What the McKinsey AI Interview Actually Is
The McKinsey AI interview is not a chatbot that asks you brainteasers. Think of it less like a trivia game and more like a structured debrief with a very patient, very attentive senior partner who never interrupts but notices everything. Lilli presents you with a business scenario, asks you to walk through your thinking in real time, and evaluates the quality of your reasoning, not just your final answer.
The format uses natural language processing to assess structure, logic, and communication clarity. McKinsey is testing whether your thinking is organized, whether you can handle ambiguity without spiraling, and whether you communicate like someone who could sit in front of a client on day one.
This is not a pass-fail trivia screen. It is a structured reasoning assessment delivered through a conversational AI interface.
What the AI Is Actually Scoring You On
Pretend you are narrating a chess match to someone who cannot see the board. If you skip steps, assume shared context, or use vague language, the listener is lost immediately. That is exactly what Lilli is doing: it is checking whether your verbal reasoning holds up when the listener has no patience for ambiguity.
Specifically, the AI evaluates three dimensions. First, it checks whether your problem structuring is MECE: no double-counting, no missing branches. Second, it listens for logical connectives: do your conclusions follow from your evidence, or do you just assert things? Third, it scores communication economy: can you say the key insight in one sentence before expanding?
These are the same things a human partner evaluates. The AI just does it consistently across every candidate.
How It Differs from a Traditional Case Interview
In a traditional case, your interviewer is a human who adjusts. If you stumble on a transition, a good interviewer will ask a clarifying question that gives you a lifeline. The AI does not offer lifelines.
Imagine driving a car with no power steering. The road is the same. The destination is the same. But every correction you make requires more deliberate effort, and small drifts compound faster. That is the AI interview. Your structure has to be cleaner upfront because no one will steer you back mid-case.
The upside: the AI is also completely neutral. It has no memory of your handshake, your nervousness at the start, or whether you made eye contact. It scores the words and the logic.
Practice this framework on a real case: the adidas-yeezy-2022 case on BoardroomIQ puts you in the room. Work through it out loud, record yourself, and listen for the moments where your logic chain breaks or you assume context the listener does not have.
Practice this framework
Work through the Adidas / Yeezy 2022: When a Brand Becomes a Liability case with AI coaching.
What Structured Thinking Looks Like Out Loud
Most candidates structure their thinking on paper and then narrate loosely. That gap is what kills scores in an AI interview.
A strong AI interview response sounds like a well-labeled map. You say where you are going before you go there. You name your framework before you apply it. You signal transitions explicitly: "I will look at this through two lenses: the demand side and the cost side. Starting with demand..." That level of narration feels slightly formal in conversation but reads as precisely right to a structured reasoning evaluator.
Practice saying your issue tree out loud before you write it down. If you cannot say it clearly, you do not actually have a clear structure yet.
How to Practice the McKinsey AI Interview Before Your Interviews
Solo verbal structuring drills. Take any business prompt and set a two-minute timer. Speak your complete issue tree aloud before writing a single word. Do this daily for two weeks. The goal is to make verbal structure feel as natural as written structure.
Record and review. Use your phone to record a five-minute case walkthrough. Play it back and flag every sentence where you made an assertion without evidence, skipped a logical step, or used filler like "I think" or "kind of." Count the flags. Reduce them by half each week.
AI-simulated case pressure. Work through a full case on BoardroomIQ without pausing or rewinding. Treat each prompt as if the evaluator is live. Do not edit your answer after you give it. Train yourself to commit to structure on the first pass.
The best way to practice the McKinsey AI interview is under realistic pressure, with a case that fights back. Open a case on BoardroomIQ right now and say your first hypothesis out loud before you type a single word.