Inside McKinsey's New AI Interview (and How to Prepare)

The most important thing to understand: this is not a prompt-engineering exam. McKinsey is not checking whether you can craft a clever instruction to a chatbot.

Based on how it's been described by candidates and career offices, the round centers on a more valuable question: Can you work with AI output the way a good consultant would? That means demonstrating:

• Curiosity — probing the AI's answer rather than accepting it.
• Skepticism — assuming the output is plausibly wrong and looking for where.
• Judgment — knowing which parts to keep, which to discard, and how to adapt a generic answer to a specific client's reality.

In other words, McKinsey took the most human, hardest-to-fake part of consulting and built an interview around it. AI generates a confident answer; your job is to be the consultant who knows it isn't good enough yet.

Two forces converged.

First, AI is now woven into the actual job. McKinsey's internal assistant, Lilli, runs over 500,000 prompts a month. If consultants spend their days directing and verifying AI, it makes sense to hire people who are good at exactly that.

First-year work used to be research and slide-building — and AI now does most of it. So the firm needs a different signal at the hiring gate. The old case interview tests structuring and math. The AI round tests the layer above: judgment about machine-generated analysis.

Second, the rest of the industry is watching. Observers expect Bain and BCG to introduce similar elements. If you're recruiting in 2026 and beyond, treat the AI-augmented interview as a category, not a McKinsey quirk.

You don't need special tools. You need a repeatable practice.

Step 1 — Generate a recommendation. Take any business situation (a real one is best — a company you follow, or a case study) and ask an AI tool for a strategic recommendation. Accept its first answer at face value, briefly.

Step 2 — Attack it. Now play consultant. Ask yourself, out loud:

• What assumptions is this answer making? Are they true for this specific company?
• What data would change this recommendation entirely?
• What did it ignore — the politics, the balance sheet, the implementation reality?
• Where is it confidently generic when the situation is specific?

Step 3 — Rebuild it. State the version you would actually give a client, and explain what you changed and why. This is the muscle the interview rewards.

Step 4 — Do it on real decisions. Generic practice helps; reasoning through actual corporate decisions builds the instinct. Working through case studies where you can see how the real company decided — and where the obvious answer was wrong — trains exactly the skepticism McKinsey is screening for.

Candidates who struggle with this round tend to treat AI as an authority. Candidates who excel treat it as a very fast, very confident junior analyst — useful, fast, and never to be trusted without verification.

That reframing is the whole game. Your value isn't competing with the AI on speed or recall. It's the judgment to know when its clean answer is wrong, and the communication to explain why.

If you've prepared for traditional case interviews — structuring problems, doing mental math, driving to a recommendation — you already have the foundation. The AI round just adds one layer: do all of that to the machine's output, not from a blank page.

For the broader picture of how AI is reshaping case interviews across all firms, see our guide on passing an AI-era case interview.

---

BoardroomIQ trains the structured judgment that AI-augmented interviews reward. Practice on real corporate decisions and live case simulations at boardroomiq-ai.com.