Working with AI inside a case interview
The format, the skills assessed, and original practice exercises with worked answers.
Firm processes and online tests change from year to year and differ by office. Use this to prepare, and confirm the exact current steps on the firm's own careers page.
Key takeaways
- Treat the AI like a fast, confident junior colleague: useful, but every number it gives you is yours to check before you say it out loud.
- These exercises are original and practice the skills that public reports describe.
- The strong candidate owns the structure and uses the AI as a fast assistant, then checks it.
- Follow each firm's rules on AI: Bain's official assessment page says AI tools are strictly prohibited during its online test, including recording, screenshots, and any software that transcribes or gives feedback.
Official facts come from the firms' own pages. Reported facts come from candidate accounts, university career centres and prep sites, and they vary by office and recruiting cycle. Your invitation always wins: follow what it says. We re-check this page at least every 90 days.
A new format is appearing in 2026: case interviews where you are allowed, or asked, to use an AI tool while you solve the problem. The interviewer watches how you use it. The goal is not to test clever prompt writing. It is to see whether you can frame the problem, check what the AI gives you, catch its mistakes, and turn the output into your own clear recommendation.
What firm leaders have said (as reported)
- McKinsey launched a free AI practice tool in April 2026 for applicants to entry-level Business Analyst and Associate roles, available globally to candidates invited to interview through McKinsey's candidate preparation site. It lets them practice case math as often as they like. McKinsey's talent attraction lead said the aim is to give every candidate free access to preparation (Business Insider, 24 May 2026). The tool is at prep-partner.mckinsey.com.
- The same lead told CaseCoach (22 September 2026) that McKinsey's AI-assisted case is about as long as a normal case interview, usually replaces one of the usual cases, and is not a standard feature worldwide; candidates are more likely to meet it in European offices. Where it is used, performance counts toward the decision to move a candidate forward or make an offer. The lead said McKinsey looks at what people do with AI and how they then shape their own answer.
- The same lead told Business Insider (24 May 2026) that the interview component asks candidates to use a publicly available generative AI tool to work through hypothetical client problems.
- Bain's global head of talent acquisition, Ron Kermisch, has said candidates should expect to use AI in the interview process. Bain has not published the format (CaseCoach, 22 September 2026).
- BCG's global recruiting executive director, Adele Lomax, has said AI aptitude is one input of many, not a stand-alone filter (CaseCoach, 22 September 2026).
- Bain's official digital assessment page says AI tools are strictly prohibited during that online test. Rules for a firm's online test and for its interviews can differ.
Reported: how the pilots run (still changing)
- Earlier, unconfirmed reports: a prep company, repeated by Fortune in January 2026, said McKinsey was piloting a final-round AI interview for US and North American candidates using Lilli, its internal AI tool. McKinsey did not respond to Fortune, and its later on-record statements above describe a publicly available tool and mainly European offices.
- BCG is reported to be piloting several approaches in select offices: live cases with access to a large language model, questions about your AI experience, and online scenario assessments.
- Bain was reported to be introducing an AI-enabled interview component from summer 2026, in which candidates use an internal Bain tool while the interviewer observes.
What is assessed
- Framing: turning the client's question into a clear, specific request with context and limits.
- Checking: testing the AI's numbers and claims against the data you have and against common sense.
- Spotting errors: units, time periods, percent versus percentage points, invented facts, and missing steps.
- Judgement: deciding what to give the AI and what to do yourself, and when to stop using it.
- Synthesis: turning the output into your own answer, with a clear so-what and next step.
- Care with data: never paste confidential information into a tool you were not given.
The interviewer says: "Use the AI tool to help size the market for a home coffee subscription in one country."
Weaker answer
Types "What is the market size for coffee subscriptions?", reads the number out, and cannot explain where it came from.
Stronger answer
Types a framed request: "Estimate the yearly revenue pool for home coffee subscriptions in this country. Use households, the share who drink coffee at home, the share who might subscribe, and yearly spend per subscriber. Show each assumption and the math." Then checks each assumption aloud, fixes one that looks too high, and states the final number with a so-what.
Why the stronger answer wins: The strong candidate owns the structure and uses the AI as a fast assistant, then checks it. The weak candidate hands over the thinking and cannot defend the answer.
Practice exercises (original, written for this platform)
An AI tool writes: "50 million households times 30 percent who might buy times 20 dollars per month gives a market of 300 million dollars a year." Find the error. What is the correct yearly market, in millions of dollars?
The AI writes: "Operating margin rose from 10 percent to 12 percent, a 2 percent increase." What is the correct relative increase in the margin, in percent?
The AI summarises: "Store A earns 1.2 million dollars on 400 square metres. Store B earns 1.5 million dollars on 600 square metres. So Store B is more productive." How many more dollars of revenue per square metre does Store A earn than Store B?
Framing exercise: a client asks why its online sales fell last quarter. Before you type anything into an AI tool, write the request you would give it, and list two things you would check in its answer.
Key idea
Treat the AI like a fast, confident junior colleague: useful, but every number it gives you is yours to check before you say it out loud.
Bain's official assessment page says AI tools are strictly prohibited during its online test, including recording, screenshots, and any software that transcribes or gives feedback. Meanwhile, some firms now pilot interviews where you are asked to use an AI tool the firm provides. Read every invitation closely, and use outside AI help only when the firm clearly allows it.
These exercises are original and practice the skills that public reports describe. They do not copy any firm's AI interview. So What Club is not affiliated with or endorsed by any firm or test provider named here.
Sources, official first (checked 28 September 2026)
- McKinsey AI Prep Tool (official, free; opens for candidates invited to interview): https://prep-partner.mckinsey.com/
- Bain careers, digital assessment (official): https://www.bain.com/careers/hiring-process/digital-assessment/
- Business Insider via Yahoo Finance, McKinsey wants candidates to practice with AI (24 May 2026): https://finance.yahoo.com/sectors/technology/articles/mckinsey-wants-job-candidates-practice-095201522.html
- CaseCoach, How McKinsey, BCG and Bain are testing AI skills in interviews (22 September 2026): https://casecoach.com/b/how-mckinsey-bcg-and-bain-are-testing-ai-skills-in-interviews/
- Fortune, how to get hired at McKinsey (14 January 2026): https://fortune.com/2026/01/14/how-to-get-hired-at-mckinsey-ai-tools-liberal-arts-creativity/
- Management Consulted, Bain AI interview: https://managementconsulted.com/bain-ai-interview/
Sources for this lesson (6)
- McKinsey AI Prep Tool (official, free; opens for candidates invited to interview)
- Bain careers, digital assessment (official)
- Business Insider via Yahoo Finance, McKinsey wants candidates to practice with AI (24 May 2026)
- CaseCoach, How McKinsey, BCG and Bain are testing AI skills in interviews (22 September 2026)
- Fortune, how to get hired at McKinsey (14 January 2026)
- Management Consulted, Bain AI interview
My notes on this lesson
0 of 5,000 characters. Saves automatically.
Try the 4 remaining checks and drills above to complete this lesson (0 of 4 done).
Spotted something wrong or out of date? Report a mistake. We check every report and correct the page.