AI case interview practice: what it’s good for, and what it isn’t
In this guide
What an AI case interviewer is
An AI case interviewer plays the interviewer’s side of a case. It opens with the client’s problem, asks one question at a time, hands over an exhibit when you ask for the data, and reacts to what you actually say: it probes a structure with a gap in it, questions an assumption, and asks you to commit to a recommendation. At the end it marks your answers against what a strong candidate would have said.
That is different from a case book, where you read the question and then the answer, and from a quiz, where you pick from options. The point of a case interview is the conversation: structuring out loud, doing the maths with someone watching, and changing course when new facts arrive. If you are new to the format, start with what a case interview is, then come back.
Try it: three questions with Ace
Ace is the interviewer in Case Ace. Below is a short exchange from AquaFresh, the sample case in our library: a laundry chain losing customers to start-ups. Pick an answer at each turn and Ace answers what you picked, in his own voice. In a live case you type or speak your own answers instead of picking.
| Service | Capex ($) | Revenue, YoY | Profit margin |
|---|---|---|---|
| Tech assets (app, website) | 200,000 | +3% | 0% |
| Free pick-up and drop-off | 100,000 | +10% | −2% |
| Clothing sanitisation | 300,000 | +3% | +2% |
| Shoe shine | 500,000 | +4% | +3% |
- Ace: AquaFresh runs a chain of laundries and is losing customers to new start-ups. The owners want revenue to grow at least 10% a year. Here are four services the start-ups offer, and what each would do for AquaFresh. Which should it prioritise?
- You: Free pick-up and drop-off. It’s the only one that reaches 10%.
- Ace: Right, it’s the only one that clears 10%.
- Ace: But look at its margin: minus 2%. Are you comfortable recommending a service that loses money?
- You: Yes, if pick-up customers spend more on other services. I’d test that, and charge for faster delivery.
- Ace: Good. That’s the real test: does it pay back through the rest of the business.
- Ace: The owners have thirty seconds. What do you tell them?
- You: Launch free pick-up and drop-off. It’s the only service that reaches 10% growth. It loses 2% at first, so pilot it in a few outlets and check that pick-up customers spend more elsewhere.
- Ace: That’s how to close: the answer first, then the reasons, then the risk and the next step.
- Ace: That’s the end of the taster. In a real case, you speak to me live, or type, and answer however you like. I respond to what you actually say. You choose how I interview you: Helper steers you when you’re stuck, Fair runs it like a real interviewer, and Adversary pushes hard on every assumption. Case Ace gives you 10 full cases from the library, 45 lessons, 9 kinds of marked drill, and a report on five skills after every case. Looking forward to interviewing with you!
What AI practice does well
Volume, whenever you have an hour
Cases get easier with repetition, and finding a partner for every one of them is the hard part: schedules clash, and a partner who is also preparing can only give you so many evenings. An AI interviewer is there at 7am before a lecture or at midnight after a shift, so the number of cases you run depends on your time, not someone else’s.
Pushback that doesn’t go easy on you
A friend playing interviewer tends to accept a decent answer and move on. A well-built AI interviewer is set up to test it: it asks what your structure leaves out, whether your number is sensible, and what you would tell the client. Case Ace lets you choose how hard it pushes: Helper steers you when you’re stuck, Fair runs it like a typical interviewer, and Adversary pushes on every assumption.
Exhibits handed over the way they are in the room
In a real case, the chart arrives when you ask for data, not before. Practising with the exhibit withheld trains the habit of asking for the right information, and of saying what the chart means for the client before you read out its numbers.
The same marking every time
Human feedback varies with the person giving it. An AI interviewer marks every case against the same criteria, so you can see whether you are improving. Case Ace marks five skills out of 5 after every case: Structuring, Quant, Judgment, Synthesis and Adaptability, with each question reviewed against an ideal answer.
Speaking practice
Typing an answer is easier than saying it. Answering out loud, with the interviewer speaking back, is closer to the real thing, and it shows you where you ramble. You can switch between typing and speaking in Case Ace.
Where it falls short
- It can’t read the room. A partner at a firm notices when you are flustered, when you have lost the thread, or when your tone undersells a good answer. An AI marks what you said, not how the interviewer felt hearing it.
- It isn’t a firm’s house style. McKinsey runs interviewer-led cases, while Bain and BCG cases are often more candidate-led, and each interviewer has habits. A coach who interviewed at a firm can tell you what that firm’s partners listen for.
- The fit interview needs a person. Your stories about leadership, impact and why consulting land differently with a human, and the follow-up questions are less predictable. Practise those with people.
- Marks are a guide, not a verdict. A high score on a practice case says you hit what the marking looks for. It doesn’t promise an offer, so treat a run of good marks as a sign you are ready for harder cases and real mocks.
- Paper maths. In most interviews you calculate on paper, in front of someone. Do the maths by hand when you practise with an AI, too.
Practising with a general chatbot
You can run a case with a general-purpose chatbot such as ChatGPT if you set it up carefully. Give it the full case, including the data and the answer, and tell it to play an interviewer: ask one question at a time, hold back each exhibit until you ask for data, push back when your structure has a gap, and mark you at the end against criteria you give it.
Without that set-up, a general chatbot tends to answer the case for you, share all the data at once and accept weak answers. It also has no exhibits of its own unless you supply them, and its marking can drift from one session to the next. If you go this route, write the case yourself from a case book and keep your own score sheet so your progress is comparable.
AI in the real process: BCG’s Casey
AI isn’t only a practice tool. BCG’s online case runs through a chatbot called Casey: one case of about 8 to 10 questions in 25 to 35 minutes, then a short video recommendation. The questions mix multiple choice, numeric answers and short written responses rather than a free conversation, so it tests different things from a live case. Read the BCG Casey chatbot guide for the format and how to prepare.
How to combine AI, partners and coaches
Each kind of practice covers what the others miss. A typical month before first-round interviews might look like this:
| Practice | How often | What it’s for |
|---|---|---|
| Case drills | Most days, 15 minutes | Maths, charts, structures and brainstorms, one skill at a time |
| AI cases | 2 to 3 a week | Full cases with pushback and marks, at any hour |
| Case partner | 1 to 2 a week | Live back-and-forth, giving cases as well as taking them |
| Coach | 1 to 3 sessions, before the rounds | Firm style, presence and the fit interview |
Start with drills and a few AI cases until the structure of a case feels routine. Then add a practice partner and take turns giving cases, since playing interviewer teaches you what a good answer sounds like. Bring in a coach once your cases are steady, to work on what an AI and a peer can’t judge: browse former consultants in the coach directory. Keep the AI cases going throughout, as your volume and your marks over time. For the drills, see Case Drills.
Frequently asked questions
Can AI replace practising with a person?
No. AI is good for volume, consistent pushback and marking, but a person is better at judging presence, the fit interview and a firm’s style. Use both: AI cases for repetition, partners and a coach for live practice.
How many practice cases should I do before the interview?
There is no fixed number. Candidates often aim for a few dozen across a month or two. Aim for the point where structuring and case maths feel routine and your marks hold steady on harder cases, rather than a count.
Is practising with ChatGPT good enough?
It can help if you set it up as an interviewer: give it the case and the data, and tell it to ask one question at a time, withhold exhibits and mark you against set criteria. Without that set-up it tends to solve the case for you and accept weak answers.
Do consulting firms use AI in their interviews?
BCG’s online case uses a chatbot called Casey, with multiple-choice, numeric and short written questions and a short video recommendation. Live case interviews at McKinsey, BCG and Bain are still run by people.
Should I type or speak my answers to an AI interviewer?
Speak them when you can. Real interviews are spoken, and saying an answer out loud shows you where it is long or unclear. Typing is useful when you can’t talk aloud, or when you want to slow down and get a structure right.
Can I try Case Ace for free?
Yes. A free account includes a Case Ace drill that ends on an example report, three lessons and every case brief. Case Ace itself includes 10 live cases with the AI interviewer.
Sources (3)
- McKinsey Careers, interviewing guidance: how the firm runs its case interviews and what interviewers look for. mckinsey.com
- BCG Careers, interview process overview: the online case with the Casey chatbot and the live case interviews after it. bcg.com
- Bain Careers, interview preparation guidance: the candidate-led case interview format. bain.com