Yes, case interview practice with AI works for most of what a case tests: give an AI chat a clear set of interviewer rules and it can run full interviewer-led or candidate-led cases, release data only when you ask for it, drill your case math and push you to a final recommendation. It is weaker at realistic pressure, consistent numbers and nuanced feedback, so pair it with a few live cases with a person.
This guide gives you the prompts we wrote for both case formats, a table of starting structures, math drills, a short sample case exchange, and an honest list of the limits. If you are new to the format, start with what a case interview is, and see our guide to the types of job interviews for how cases compare with other rounds.
What parts of a case interview can you practice with AI?
| Case skill | How well AI practice works | How to practice it |
|---|---|---|
| Structuring the problem | Well | Ask for a case prompt, give your structure out loud, then ask for a critique of gaps and overlaps |
| Case math | Well, if you check its arithmetic | Timed drills on percentages, margins, breakeven and large numbers |
| Interpreting data | Reasonably | Ask it to describe a chart as a table and make you state the "so what" |
| Brainstorming | Well | Ask for a brainstorm question, then for ideas you missed |
| Synthesis and recommendation | Well | Give a 60-second recommendation and ask whether it answers the original question |
| Driving a candidate-led case | Reasonably | Tell it to answer only what you ask and never suggest the next step |
| Composure under pressure | Poorly | Live cases with a person |
| Fit and behavioral questions | Well | Scored mock interview answered out loud |
Speak your answers, even when you type them afterwards. Case interviews reward thinking out loud, and typed answers let you hide pauses and false starts that a real interviewer would hear.
How do you run an interviewer-led case with AI?
In an interviewer-led case, the interviewer moves you through a set sequence of questions. That suits AI well, because the sequence can be planned up front. Paste this prompt into any AI chat.
Prompt we wrote (interviewer-led case):
You are a case interviewer at a strategy consulting firm. Before we start, write a complete fact sheet for a [profitability / market entry / pricing] case about a [industry] client, keep it to yourself, and keep every number consistent with it. Then: 1) Give me only the case prompt. 2) Wait for my structure and react in one or two sentences, as a real interviewer would. 3) Ask me three or four questions in order: one calculation, one table of data I must interpret, one brainstorm, and a final recommendation. Give data only when I ask a relevant question. Never do the math for me, but tell me if my arithmetic is wrong. 4) At the end, score me 1 to 5 on structure, math, insight, synthesis and communication, with one specific fix for each.
Ask for a clarifying minute before you structure, just as you would in a real case. If the AI's reaction to your structure is pure praise, add a line to the prompt: "Be as demanding as a senior interviewer; point out at least one weakness each time."
How do you run a candidate-led case with AI?
In a candidate-led case, you drive: you decide which branch of your structure to explore, ask for data, and move on when you have what you need. The main risk with AI is that it helpfully drives for you. Shut that off explicitly.
Prompt we wrote (candidate-led case):
Run a candidate-led [market entry] case. Write the full fact sheet first, keep it to yourself and keep it consistent. Give me the case prompt, then answer only the questions I ask, using only the fact sheet. Do not suggest what to look at next. If I ask for data that does not exist, say "We don't have that; what would you assume?" If I go two turns without making progress, ask "Where would you like to go next?" When I give a recommendation, ask me two pushback questions, then give feedback on how well I prioritized.
Formats vary by firm and even by interviewer, so practice both styles. Many consulting firms publish sample cases on their careers sites; work through those as well, because they show the style that firm uses.
Starting structures and case math drills
Common advice from consulting interviewers is to build a structure that fits the specific client rather than reciting a memorized framework. The buckets below are starting points to adapt, not scripts.
| Case type | Starting buckets to adapt | Good early question |
|---|---|---|
| Profitability | Revenue (volume, price, mix) and costs (fixed, variable) | Is the decline in revenue, costs, or both? |
| Market entry | Market attractiveness, competition, client capabilities, entry options, economics | What does the client want from entering: growth, scale, defense? |
| Pricing | Cost floor, customer value, competitor prices, strategic goal | Is this a new product or a price change? |
| Acquisition | Target's standalone value, synergies, price and risks, fit with strategy | Why does the client want to acquire now? |
| Growth | Existing customers, new customers, new products, new markets | What growth target, and over what period? |
| Market sizing | Population or units, share that buys, frequency, price | Are we sizing by revenue or by units? |
For math, ask for drills rather than full cases, and time yourself:
Give me 10 case-math questions one at a time, mixing percentage change, margins, breakeven, growth over several years and multiplication of large numbers. Wait for my answer before showing the correct one, then show the fastest mental method. Keep a tally and list the question types I got wrong at the end.
Check any answer that looks off with a calculator. The goal is speed with clean notation: write units, keep zeros organized, and say each step out loud.
A sample case exchange we wrote
This is a shortened interviewer-led profitability case, written by us to show the rhythm you want to practice.
- Interviewer: Your client runs 40 gyms. Revenue has been flat at $60 million for two years, but profit has fallen 20%, from $9 million to $7.2 million. Why, and what should they do?
- Candidate: Since revenue is flat, I'll start with costs, split into staff, rent, marketing and utilities and other. I'll also check whether flat revenue hides a shift in mix, such as more discounted memberships. Can I see costs for both years?
- Interviewer: Staff went from $24 million to $24.6 million, rent from $15 million to $15.3 million, marketing stayed at $6 million, and utilities and other went from $6 million to $6.9 million.
- Candidate: Total costs rose $1.8 million, which matches the profit decline. Utilities and other is $0.9 million of that, half the problem, and a 15% increase. Staff is $0.6 million, up 2.5%. I'd focus on utilities first: is it higher usage or a higher rate?
- Interviewer: Usage is flat. Their energy contract renewed at a higher rate last year.
- Candidate: Then my recommendation is to tackle energy cost first: renegotiate or competitively bid the contract, and reduce usage with equipment scheduling and efficient lighting. Second, review staffing levels against peak hours, since staff is the next largest increase. Before acting, I'd confirm that revenue mix has not shifted toward lower-margin memberships.
Notice what the candidate does: states the structure, asks for specific data, does the math out loud, prioritizes by size, and ends with a recommendation plus a risk. Ask the AI to score exactly those moves.
What are the limits of AI for case interview practice?
- Arithmetic mistakes. Chatbots can get sums wrong or change a number halfway through. The fact-sheet instruction helps; checking with a calculator afterward helps more.
- Feedback that is too kind. Default AI feedback leans positive. Ask explicitly for weaknesses.
- No real exhibits. Text tables are not the same as reading a crowded chart quickly. Practice with published sample cases that include exhibits.
- No real pressure. An AI never sighs, interrupts or looks at the clock. Do several live cases with a classmate, colleague or case partner before the real thing.
- It cannot tell you which firm style to expect. Use each firm's own materials and your recruiter for that.
Use AI to prepare, not during the interview itself: follow each employer's rules. Consulting interviews commonly include fit questions alongside the case, and finance or consulting final rounds may stack several interviews in one day; our guide to the superday interview covers that format.
Practice it
Cases are only part of the loop. For the fit questions, and for short product or estimation questions like those in product manager interviews, rehearse out loud with an AI mock interview built from your résumé and the job description. CoPilot Interview, an AI interview assistant for practice and prep, includes a Mock Interview mode that reads each question aloud, listens to your spoken answer, scores it out of 10 with what worked, what to fix and a model answer, and ends with a report card. It is free for 15 minutes a day, with no account. It runs question-by-question rounds rather than a long interactive case, so keep the prompts above for full cases.
Build your fit stories with our guide to a STAR story bank with AI, and see how to practice with an AI interview assistant for a full weekly routine. New to these tools? Start with what an AI interview assistant is.
Rehearse your fit answers out loud
Answer behavioral and product questions drawn from your résumé and the job description, and see each answer scored out of 10 with what to fix.
Try a free AI mock interviewFAQ
Can you practice case interviews with ChatGPT or other AI chatbots?
Yes. A general AI chatbot can act as the interviewer if you give it clear rules: release data only when asked, do not solve the case for you, check your math, and score you at the end. It works well for structure, math and synthesis practice, and less well for realistic pressure and nuanced feedback.
What is the difference between an interviewer-led and a candidate-led case?
In an interviewer-led case, the interviewer steers you through a set sequence of questions. In a candidate-led case, you decide which areas to explore and ask for the data you need. Formats vary by firm and even by interviewer, so practice both.
Is AI reliable for case interview math?
Not fully. AI chatbots can make arithmetic mistakes and can change numbers mid-case. Ask it to write the full fact sheet first and keep to it, and check any figure that looks wrong with a calculator after the case.
How many practice cases should I do with AI before a consulting interview?
There is no fixed number. Use AI for many short drills and full cases, and track whether your structure, math speed and final recommendation improve. Add live cases with a person before the real interview, because human pressure is hard to simulate.
Should I memorize case frameworks?
Learn the common starting buckets, such as revenue and costs for profitability, but build a structure specific to each client. Interviewers commonly notice when a candidate recites a generic framework that does not fit the question.
Can AI help with the fit part of a consulting interview?
Yes. Fit or behavioral questions are usually ordinary interview questions, so you can rehearse them in an AI mock interview built from your résumé and the job description and get each spoken answer scored.