Generic case books take you to mediocre. A good prompt forces you to be MECE, time-box, and surface a recommendation with a clear next step instead of guessing. And the bar moved in 2026: McKinsey now runs an AI round where you use its internal assistant Lilli on a live case, and interviewers grade how you prompt it, challenge its output, and fold it into the client’s situation (per the Irish Times, Jan 2026). Bain has signaled its own AI-enabled interview for Summer 2026. So practicing with AI is no longer a shortcut — it is closer to the real thing.
TL;DR
- 12 copy-ready prompts below cover the full MBB and PM loop: market sizing, profitability, M&A, product launch, MECE, synthesis, chart reading, debrief, and pacing.
- Use Claude (Opus 4.7 or Sonnet 4.6) for full-case context and sensitivity-chain math; it tracks assumptions across a 20-minute case better than most. Use ChatGPT (GPT-5.5, Thinking mode) when you want a tougher, more adversarial interviewer.
- The new skill being tested (McKinsey Lilli, Bain AI round) is collaborating with AI under pressure — see the dedicated prompt in section 13.
- AI softens criticism by default. Every prompt here explicitly tells it to push back; do not skip that line.
Who this is for
MBB candidates (McKinsey, BCG, Bain), product-strategy and BizOps interviews, PM loops at Meta/Stripe/Google, and ex-consultants reactivating the muscle. If you have a final round in the next two weeks, jump to sections 1, 8, and 13.
Which AI model for case prep (as of June 2026)
The job is long-context reasoning plus clean arithmetic, so model choice matters more here than for most prompt work.
| Model | Plan / price | Why it fits case prep |
|---|---|---|
| Claude Opus 4.7 / Sonnet 4.6 | Pro $20/mo (Sonnet on Free, limited) | 1M-token context holds a whole case; strong, traceable math on sensitivity chains |
| ChatGPT GPT-5.5 (Thinking) | Plus $20/mo (GPT-5.5 on Free, tight limits) | Most willing to play a hard, interrupting interviewer; good curveballs |
| Gemini 3.1 Pro | Google AI Pro $19.99/mo | 1M context, strong on chart/data reading and quick sanity checks |
All three reliably run a 20-minute mock for free-tier or $20-tier accounts. For most people the call is Claude for the full-case feel, ChatGPT when you specifically want to be grilled. If you are new to either, see our ChatGPT beginner guide and Claude beginner guide.
Prompt anatomy: six elements
Every case-prep prompt should carry these. Miss one and the output drifts generic.
- Role: candidate, interviewer, recruiter — name the persona the model plays.
- Context: target firm, role, level, your background.
- Goal: one deliverable — analysis, script, answer, or critique.
- Constraints: time box, banned phrases, must-include numbers.
- Tone: how the interviewer behaves — interrupts, stays silent, pushes on math.
- Examples: paste one of your past answers so it grades against your real voice.
12 copy-ready prompt templates
Swap the backtick [placeholders] for your own values before sending.
1. Live case simulation
You are a McKinsey case interviewer. Run a profitability case for me: pose the
prompt, expect me to clarify, push back when I jump to conclusions, surface a
curveball mid-case. Do not solve it for me. Interrupt if I ramble past 30
seconds. Score the case at the end on structure, math, and synthesis (1-5 each).2. Market sizing in 5 minutes
Market sizing prompt: `[prompt]`. Before I do any math, push back on my structure
and make me justify each assumption. Then walk the clean estimate: (1) clarify,
(2) framework, (3) numbers with explicit assumptions, (4) sanity check via a
second method, (5) final answer with a confidence range.Swap: [prompt]
3. MECE framework drill
Topic: `[topic]`. Build 3 MECE frameworks: (a) revenue / cost, (b) customer /
product / channel, (c) internal / external. For each: when to use it, when not,
and one overlap trap that breaks the "mutually exclusive" rule.Swap: [topic]
4. Profitability case structure
Profitability case for `[industry]` declining margin. Structure: (1) revenue side:
volume x price x mix, (2) cost side: variable / fixed / one-time, (3) competitive
and market side. For each branch, name the one hypothesis to test first and the
data you would request.Swap: [industry]
5. M&A case
M&A case: should `[acquirer]` buy `[target]`? Structure: (1) strategic fit,
(2) financial fit (synergies, valuation), (3) integration risk, (4) alternatives.
Walk through with explicit assumptions and a go / no-go recommendation.Swap: [acquirer], [target]
6. Product launch decision
Should `[company]` launch `[product]`? Build: (1) market sizing, (2) competitive
landscape, (3) financial case (build-vs-buy, margin), (4) execution risk,
(5) recommendation with a deadline and the first 30-day move.Swap: [company], [product]
7. Curveball drill
You are an interviewer. I am mid-case. Throw a curveball — a fact I did not
expect (a regulatory change, a new competitor, a currency move). Watch whether I
re-anchor without panicking. Then critique how I integrated it.8. Synthesis under time pressure
I have 60 seconds to synthesize my case answer. Give me a structure:
(1) recommendation first, (2) 2-3 reasons, (3) one risk plus mitigation,
(4) next step. Time me. Cut anything past 60 seconds and tell me where I bloated.9. Numerical chart reading
Here is a chart: `[chart description]`. Walk through: (1) what the chart shows,
(2) one calculation worth doing, (3) one insight, (4) what extra data would
sharpen the conclusion. Keep the math clean and visible.Swap: [chart description]
10. Case debrief
After this case, debrief me: (1) Where did my framework break? (2) Did I clarify
enough up front? (3) Did I drive or follow? (4) Math accuracy. (5) Synthesis
clarity. Score each 1-5 and give me the single highest-leverage fix.11. PM product case (Stripe / Meta style)
PM product case: "Design `[product]`". Structure: (1) clarify goal, audience,
success metric, (2) user segments, (3) 3 jobs-to-be-done, (4) 3 product concepts,
(5) pick one and define how to measure success.Swap: [product]
12. Pacing and filler audit
Here is my case answer transcript: `[transcript]`. Flag: (a) sentences over 25
words, (b) restarts, (c) hedge words, (d) math hesitations. Output a coaching
summary with the 3 habits to fix first.Swap: [transcript]
13. AI-collaboration round (McKinsey Lilli / Bain style)
New for 2026: the firm hands you an AI assistant and grades how you work with it. McKinsey’s Lilli round and Bain’s planned Summer 2026 format reward curiosity and judgment, not slick prompts. Rehearse it directly.
You are an AI assistant in a consulting interview. I will give you a client case.
Answer my prompts as a helpful but sometimes flawed assistant — occasionally give
a plausible-but-wrong number or a weak structure. After each of my prompts, an
interviewer (also you) scores: did I challenge your weak outputs, re-prompt
sharply, and tie the answer to the client's specific situation? Case: `[case]`.Swap: [case]
Common mistakes
- No specific context (firm, role, level): output is generic and interviewers can tell.
- Asking the model to “be honest” without pasting your real record: it confabulates.
- One answer reused across firms: McKinsey, BCG, and Bain interviewers compare notes.
- Letting the AI agree with you: without an explicit “push back” line it softens criticism.
- Skipping fact-checks: models still invent dates, market figures, and titles.
- Treating the first draft as final: first drafts read AI-flavored.
- Deferring to the AI in the Lilli-style round: the test is whether you challenge it, not whether you can avoid it.
How to push results further
- Paste a real past answer so the model grades against your voice, not a template.
- Have it play interviewer first; weak answers expose themselves fast.
- Run the case live with the timer on — fatigue and pacing are half the score.
- Read every synthesis aloud; if it does not sound like you out loud, rewrite it.
- Keep a phrase bank of clarifying questions and synthesis openers that landed.
- Still run real cases with a partner. AI will not apply the same pressure a human does (see Hacking the Case Interview).
FAQ
- Which model is best for case math?: Claude Opus 4.7 or Sonnet 4.6 hold a full case in context and keep sensitivity-chain math traceable. ChatGPT GPT-5.5 in Thinking mode is the most adversarial interviewer. As of June 2026 both work on free or $20 tiers.
- Do firms care that I used AI to prepare?: No. McKinsey, BCG, and Bain only care that you can structure, do math, and show judgment live. McKinsey now requires AI use in one round, so practicing with it is aligned with the format.
- What is the McKinsey Lilli interview?: A round where you use McKinsey’s internal assistant on a live case; interviewers grade how you prompt it, judge its output, and apply it to the client. Section 13 rehearses exactly this.
- Will AI push back as hard as a real interviewer?: No. Both Claude and ChatGPT soften criticism unless told otherwise, which is why every prompt here includes an explicit push-back instruction. Keep a human partner in the loop for final rounds.
- How much research per company is enough?: 60-90 minutes for a target firm; returns diminish after. Re-scan news and launches the morning of the interview.
- How do I keep prep notes organized?: One doc per firm — research, questions to ask, story bank, and the frameworks that fit that firm’s case style.