Not long ago, practicing a case interview without a partner meant pasting a casebook case into ChatGPT and hoping it would behave. Now there are more than a dozen AI case interview tools, and on their home pages they look almost the same: an AI interviewer, a library of McKinsey, BCG and Bain style cases, a score at the end. The differences show up around minute five, when the first exhibit should land on your screen, and again around minute twenty five, when a short mock would already have ended. Those are the minutes that decide the real interview.
A disclosure first. I build Tough Tongue AI, one of the tools on this list. This year I sat down with close friends from Kellogg who now work at McKinsey, Bain and BCG, and we built case interviews together, so I have strong opinions about what realistic means here. I have described every other tool from its own website, and where a vendor does not publish something, I say so rather than guess. Prices and libraries change quickly in this category, so check anything decisive on the vendor’s site.
Below: what makes an AI case interview feel real, the tools one by one, a comparison table, and how to pick. If you would rather watch a full case before reading about one, here is a complete MBB mock on Tough Tongue, exhibits and all.
What makes an AI case interview feel real
A first round at McKinsey, BCG or Bain runs twenty five to forty minutes. You get a prompt, ask clarifying questions and lay out a structure. Then the data arrives as exhibits, a chart or a table, and the interviewer asks what you see. There is usually a second calculation, a brainstorm, and a recommendation the interviewer pushes back on to see whether you hold your ground. Practice transfers when it reproduces that shape. I use five tests, in the order a case runs into them.
Does the interviewer run the room? A voice you talk to, that asks one question at a time, waits through your silence when you ask for a minute, and probes a correct answer as hard as a wrong one. Typing into a chat skips the part where you think out loud while someone listens.
Is the exhibit on your screen? Charts are where candidates separate: reading the footnote, converting units, noticing the bump in a curve. A tool that reads the chart out to you has done the hard part for you.
Does the case run its full length? Candidates rarely fail in minute five. They fail when the second calculation lands on a shaky framework, or when the interviewer asks “why not the other market?” in the last three minutes. A ten minute case never gets there.
Is the feedback calibrated? Every casebook case comes with an answer key, and firms score against a rubric. “Good structure, clear communication” is not feedback. “Your landed cost was 31.8, the key says 31.1, you put a tariff on parts that were never imported” is.
Where does the interview happen? A lot of first rounds are now video calls. Rehearsing in a browser tab is fine for volume. Rehearsing in the same Zoom or Google Meet window, with a face on the other side, removes one more thing that feels new on the day.
Two practical questions sit on top of those: whether the cases are interviewer-led (the McKinsey format, where the interviewer holds the question order) or candidate-led (more common at BCG and Bain, where you drive), and what it costs to run thirty of them.
- 1An interviewer that runs the roomA voice, one question at a time, patience with silence, probes on right answers too.
- 2The exhibit on your screenA chart you read yourself, not a paragraph read to you.
- 3The full twenty five to forty minutesThe second calculation and the push-back live in the back half.
- 4Feedback against the answer keyYour numbers next to the key’s numbers, scored on a rubric.
- 5The room itselfA browser tab, or the Zoom or Meet window you will actually use.
- →Then: case style, library and price
The tools at a glance
Here is the short version. Everything about the other tools comes from their own websites as of October 28, 2026. The detailed feature comparison is further down.
| Tool | Best for | Try it free | Paid from |
|---|---|---|---|
| Tough Tongue AI | Exhibits on screen, full-length casebook cases, Google Meet and Zoom, building your own | 25 minutes | $20 per month |
| Road to Offer | Cases, drills and online test simulators in one subscription | 1 full graded case | $49 per month |
| The Case Room | Long voice cases with a re-answer loop | 1 case | $10 per case |
| CaseRound | Voice cases in both styles, including your own uploaded PDFs | 1 interview | $39 per month |
| MBB Prep | Written and voice AI cases, with coaching alongside | 3 AI mock interviews | $39 per month |
| Kasie | Interviewer-led and candidate-led cases while it is in beta | Free during beta | Not announced |
| CaseTutor | Voice cases in six languages | 7-day trial | $49 per month, or $249 per year |
| MBB Ready | 30-minute voice rounds with quoted evidence for every score | 1 case, partial feedback | $29 per case |
| Case Interview Labs | Cases calibrated to 11 specific firms | Half-length case | £37.99 for 3 cases |
| CaseWithAI | Pay-as-you-go cases with a coaching mode | 1 case | $1 per credit, 4 credits a case |
| CasePrepared | Voice-only cases with filler-word and think-time analysis | 3 cases | $26.99 per month |
| Case Study Prep AI | Case packs, fit interviews and Casey-style simulations | Free trial | $35 for 5 cases |
| ChatGPT, Claude or Gemini | Free structure and flow practice with a strict prompt | Yes | ChatGPT Plus is $20 per month |
1. Tough Tongue AI
Tough Tongue AI is a platform for building live AI agents for the conversations that matter most: sales calls, product demos, interviews. Case interviews are one of the things built on it, and the MBB collection is the part that matters here: 38 interviewer-led cases from the Kellogg 2024, Wharton 2024-25 and Chicago Booth 2024-25 casebooks and The Ultimate Case Interview Workbook, 20 to 40 minutes each. The case interview prep hub lists them all by casebook and difficulty. I will spend more time on it than on the others because I can describe it from the inside, and because three things about it set it apart: the slides, the meeting room, and building interviewers from the AI assistant you already use.
The interviewer drives the exhibit deck
Every case comes with a slide deck built from the casebook’s own exhibits, and the interviewer controls it. When the case reaches the data, it says “Let me pull up Exhibit A, you should see it on slide 4,” and the slide on your screen changes. The deck is gated by phase: the chart for part four stays hidden until part four begins, so you cannot read ahead. In Lux Rides, case 9 of the Wharton casebook, that matters, because the willingness-to-pay survey on slide 6 gives the price away. Ask to see an earlier exhibit again and it flips back.
This sounds like a small thing until you compare it with the alternative. An interviewer without slides has to read the exhibit out, and reading it out means choosing words, which means interpreting it for you. When we tested a strict ChatGPT and Claude interviewer prompt on a Kellogg casebook case, the narrated exhibit arrived with its conclusion already attached. Take Exhibit A in Lux Rides: a bar chart of component costs per scooter, 7.3 million dong from Spain, 13.2 from Italy and 3.5 from Vietnam, with two footnotes, shipping at 12.5% of component costs and tariffs at 20% of component costs. The trap is that the Vietnamese parts are never imported, so they carry no tariff. Here is that exhibit both ways.
It can run the interview on Google Meet or Zoom
A lot of first rounds now happen on a video call, and the call adds its own friction: the gallery view, half a second of lag, talking to a camera instead of a person. The same interviewer can join a Google Meet or Zoom meeting as a participant and run the case there. Paste the meeting link and it joins, with its voice and, if you pick one, a live avatar face. You rehearse in the window you will use on the day, with your own camera on and the meeting clock running.
The face is optional, and there is a choice of them. In the browser, each MBB case shows its interviewer as a portrait or, in newer cases such as Lux Rides, as a video that switches between listening and speaking. Anyone building a case can use a live avatar from Anam, HeyGen LiveAvatar, Avatario or Protoface instead. That is not decoration in a case interview. A good part of what makes the real one stressful is being watched while you do arithmetic, and a face that reacts gets you used to it.
A consulting club can use the same thing for a group. Several members join one call and watch each other case, which is how most club case nights already work, except that the interviewer is always available and always calibrated to the same key. Meeting deployment and live avatars are included from the $20 per month Starter plan.
You can build your own interviewer from ChatGPT, Claude, Codex or Cursor
The 38 cases are the ones we have converted so far, and the list grows every week. You do not have to wait for us, though. Tough Tongue has an MCP server, which means the AI assistant you already use can create and edit interviewers for you in plain language. There is an official Tough Tongue AI app in ChatGPT, a Claude plugin (in Claude, go to Settings, then Plugins, add a marketplace and paste https://github.com/tough-tongue/toughtongue-skills; it needs a Pro, Max, Team or Enterprise plan), and a one-line install for coding agents such as Claude Code, Codex and Cursor. Sign-in is OAuth, so there is no key to copy around. The setup for every client is on the agents page.
For casebook cases specifically, we open-sourced the workflow behind the collection as a skill, case-interview-builder, in our community skills repo. You give it a case: a page range from a casebook PDF, a doc or pasted text. It separates the prompt, the clarifying information the interviewer should reveal only when asked, the exhibits and the answer key, and it does not invent any number that is not in the source. It builds the exhibit deck in Gamma, waits for you to publish the deck to Google Slides, then creates the interviewer on Tough Tongue with the key hidden, a weighted rubric and a phase-by-phase study guide. In a coding agent, two commands install everything:
npx plugins add tough-tongue/toughtongue-skillsnpx plugins add tough-tongue/ttai-community-skillsThis is the part a consulting club should care about most. Your club’s casebook, the case an alumnus swears by, a case from the industry you are recruiting for: each can become an interviewer that every member runs the same way and gets graded against the same key. The skills are community-maintained, so look over the deck before you publish it.
- “Turn pages 42 to 51 of this casebook into an interviewer-led case interview with exhibits.”
- “Build a market sizing case for a bank entering Brazil, and have the interviewer challenge every assumption.”
- “Pull my last five case sessions and tell me which category I keep losing points on.”
The details that make the feedback calibrated
The three features above are the visible ones. These are the ones that make the practice transfer.
- It does not correct your math. Get a number wrong and it asks you to walk through the calculation until you find the error. It gives you the figure only after two failed nudges, and the report notes that it had to. An error you catch yourself is scored more leniently than one you never catch.
- It probes right answers and pushes back on your recommendation. A correct number still gets a “why?”, and most cases end with one objection built from something you said earlier.
- A conductor keeps time. In Lux Rides it nudges you to wrap up at minute 40 and closes the session at 45, whether or not you are ready.
- The report is graded against the case’s answer key. Your numbers are pulled from the transcript and set next to the key’s, each score comes with a quote of what you said, and you get a hire band, your top three improvement areas and a percentile against everyone who has run the case. Seven of the cases also come with a study guide.
Pricing
A free account gets 25 minutes to try the format, but free sessions run 10 to 12 minutes, so a full case needs a paid plan. Starter is $20 per month for 150 minutes, four or five full cases, and includes Google Meet and Zoom deployment and live avatars. Pro is $99 per month for 900 minutes that roll over. Clubs and career centers can use a team workspace with a shared pool of minutes, starting at $299 per month for 2,990 minutes and 15 seats. Details are on the pricing page.
Where it falls short
Every case in the collection is interviewer-led. Six of the nine Wharton cases are interviewee-led in the original casebook and run interviewer-led here. If you are recruiting for BCG or Bain, practice driving a case yourself as well, and use these sessions for the exhibits and the math.
An AI interviewer also cannot judge presence, whether you seem like someone a partner would put in front of a client. And the report follows the casebook’s answer key, quirks included. When you disagree with a score, the transcript and the quoted evidence are there to check.
Best for: candidates who want the closest thing to the real room, with exhibits on screen, the full length, and the option of a real Zoom or Meet call, and clubs that want to turn their own casebook into interviewers.
The other AI case interview tools
The rest of the list, described from each tool’s own site. Where a site does not say something, I leave it out rather than guess. The first seven run full AI case interviews, the next four are more specialised, and the last is the free route.
2. Road to Offer
Road to Offer is an all-in-one prep platform: 71 cases, 61 or more of them in a voice mode for answering out loud, plus guided and learning modes in text, more than 550 drills, and simulators for BCG’s Casey and CCA, McKinsey Solve and Bain’s SOVA tests. Feedback scores seven skills, from problem understanding and MECE structure through exhibit interpretation to synthesis, and quotes your own answers back to you. Its site does not say how exhibits are displayed during a voice case or how long a regular case runs; the Casey simulator is listed at 25 to 35 minutes.
Best for: candidates who also face online tests and want cases, drills and simulators in one subscription. Price: one full graded case free, then $49 per month, $120 for 90 days on a recruiting season offer, or $249 per year.
3. The Case Room
The Case Room was built by an ex-McKinsey founder. It runs voice cases of 30 to 40 minutes including the debrief, across 150 cases, with real exhibits shown on screen in its demo. After a case it shows you a stronger answer and asks you to answer again, then puts the two side by side, which is a good way to make feedback stick. You can pause mid-case, and human coaching over Zoom is available separately at $250 per 50-minute session.
Best for: candidates who learn by re-answering. Price: one case free, $10 for a single case, or one-time packs of 5 cases for $45, 15 for $109 and 30 for $199, used within 12 months.
4. CaseRound
CaseRound is a voice simulator where the interviewer reveals exhibits at the moments the case reaches them. Cases run about 40 minutes, come in both interviewer-led and candidate-led styles at three levels, and number 30 or more. It scores eight dimensions, including presence and case leadership. Paid users can also upload their own case PDFs and run them as voice interviews with exhibits.
Best for: candidates who have a casebook and want it run as voice interviews inside one app. Price: one interview free; Pro is $39 per month for 50 voice sessions, and you can cancel any time.
5. MBB Prep
MBB Prep has 50+ cases in its library, 50+ voice AI mock interviews and 500+ drills. Its interview screen shows the exhibit alongside the conversation, and it scores you on six dimensions against what it calls the rubric MBB interviews use. Coaching with its ex-Bain founder runs alongside the platform.
Best for: candidates who want written cases, voice AI cases and a human coach option in one place. Price: a free account with six AI credits at sign-up, enough for three AI mock interviews, then the Prep Pass at $39 per month, about ten mock interviews a month.
6. Kasie
Kasie is free while it is in beta. It has 20 cases and counting, in both interviewer-led and candidate-led styles, with charts and data exhibits, and aims to take you from zero to feedback in about 30 minutes. The interviewer speaks (its site calls it AI narration); whether you answer by voice is not stated. Feedback covers six areas, including case leadership, and it rewrites your weakest answers the way a top candidate would have given them.
Best for: trying a structured AI case at no cost right now. Price: free during the beta, with paid plans not yet announced.
7. CaseTutor
CaseTutor runs voice cases with timed exhibits and interviewer personas that push back, and scores structure, math, insight and communication. It lists 75+ cases, supports six languages, and covers behavioral interviews and firm-specific online tests too.
Best for: candidates who want to practice in a language other than English. Price: a seven-day free trial, then $49 per month, $99 for three months or $249 per year.
8. MBB Ready
MBB Ready runs 30-minute voice rounds with named AI interviewer personas and lets you keep the transcript and exhibit afterwards. Its rubric has three dimensions and 14 sub-dimensions on a five-point scale, and every score cites a quoted moment from your transcript. A live final-round simulation with a human partner on Zoom is sold as an add-on.
Best for: candidates who want every score tied to evidence. Price: one free case with the overall score and the start of the feedback, then $29 per case or a $149 season pass.
9. Case Interview Labs
Case Interview Labs calibrates its cases to specific firms, 11 of them, including McKinsey, BCG, Bain, Deloitte, Strategy&, Kearney and Accenture, at five levels. The AI interviewer speaks every turn and you answer by voice or keyboard, and it hands you a chart when a real interviewer would. The debrief is a scorecard against the firm’s bar at your level, with a verdict and what to practise next.
Best for: candidates interviewing beyond MBB who want the firm’s own style. Price: a free half-length case, then £37.99 for three cases, down to £8 a case on the 25-case pack, on credits that do not expire.
10. CaseWithAI
CaseWithAI shows an AI interviewer video with a live transcript, offers interviewer-led and candidate-led cases, and has a progressive coaching mode that gives feedback after every answer rather than only at the end. It has a team plan for business schools, consulting clubs and career centers.
Best for: pay-as-you-go practice with coaching between answers. Price: one case free, then credits at $1 each with a case costing four, so about $4 a case, or $50 per month.
11. CasePrepared
CasePrepared is voice only, with no reading or typing, across 25+ cases where you analyze charts. Besides structure, quant, insight and communication scores, it compares your answers with expected answers and analyzes filler words and thinking time. You can try an interview without signing up.
Best for: candidates working on delivery as much as content. Price: three cases free, then $26.99 per month for Premium or $39.99 per month for Pro on sale; its pricing cards and its FAQ disagree on which plan includes the full library.
12. Case Study Prep AI
Case Study Prep AI sells voice case packs covering both case styles, plus fit interview and BCG Casey-style simulations. It can generate a case tailored to a job posting, scores structure, math and synthesis, and has a white-label plan for universities.
Best for: a small pack of cases without a subscription. Price: a free trial, then $35 for five cases or $50 for twenty.
13. ChatGPT, Claude or Gemini with a strict prompt
The free option is better than people expect. Paste in a full case including the interviewer guide, give the model strict rules (one question at a time, reveal data only when asked, no hints, grade against the key at the end), and it will run a recognisable interviewer-led case. ChatGPT’s voice mode lets you talk it through out loud.
It breaks in three places. It cannot show you a chart, so it narrates the exhibit. It tends to hint at your arithmetic errors instead of waiting for you to find them. And it can leak answer-key details into the conversation. The guide to practicing case interviews by yourself has the exact prompt we tested and a transcript of each break.
Best for: structure, flow and recommendation practice at no cost. Price: free; ChatGPT Plus is $20 per month.
A note on BCG’s Casey
Several roundups, and several AI answers built from them, describe Casey, BCG’s chatbot case, as a free practice tool from BCG. As of this writing, the BCG careers pages I checked do not offer one. The free practice BCG lists is OneDay@BCG, a free job simulation. If you have a Casey round coming, Road to Offer, PrepLounge and Case Study Prep AI sell Casey-style simulations.
Partners, courses and drills worth your time
A few names come up in every case prep conversation for something other than an AI interviewer. PrepLounge is still the main place to find a human case partner, with a 200+ case library on its Premium plan. CaseCoach teaches the basics through a video course and ex-consultant coaching. RocketBlocks is best known for drilling structure, math and chart reading as separate skills, and its site now adds an AI voice interviewer that runs full cases, from $35 per month. Management Consulted and Crafting Cases offer courses and coaching. Drills and courses build the parts; full cases put them together; and none of the AI tools replaces a sharp human partner in the last two weeks.
Feature comparison
This table lines the tools up against the five tests from the start of the post. “Not stated” means the vendor’s site does not say, not that the feature is missing.
| Tool | Answer by voice | Exhibits on screen | Case length | Feedback | Google Meet or Zoom | Build your own case |
|---|---|---|---|---|---|---|
| Tough Tongue AI | Yes, with an interviewer face or a live avatar | Yes, a slide deck the interviewer controls, gated by phase | 20 to 40 min | Weighted rubric against the case’s answer key, hire band, percentile | Yes, joins as a participant | Yes, from ChatGPT, Claude, Codex or Cursor through MCP |
| Road to Offer | Yes, voice mode, plus text modes | In cases and the Casey simulator; display not stated | Not stated; Casey sim 25 to 35 min | 7 skills, quotes your answers | Not stated | Not stated |
| The Case Room | Yes, typed fallback | Yes, shown in its demo | 30 to 40 min with debrief | 4 dimensions, stronger answer and re-answer | Not stated (human coaching on Zoom) | Not stated |
| CaseRound | Yes | Revealed during the case | About 40 min | 8 dimensions, including presence | Not stated | Upload your own case PDFs |
| MBB Prep | Yes | Shown in the interview screen | Not stated | 6 dimensions, against an MBB-style rubric | Not stated | Not stated |
| Kasie | Not stated; the interviewer speaks | Charts and data exhibits | About 30 min | 6 dimensions, rewrites weak answers | Not stated | Not stated |
| CaseTutor | Yes | Timed exhibits | Not stated | 4 dimensions | Not stated | Not stated |
| MBB Ready | Yes, named personas | Yes, kept after it is shown | 30 min | 3 dimensions, 14 sub-dimensions, quoted evidence | Not stated (human partner on Zoom as an add-on) | Not stated |
| Case Interview Labs | Voice or keyboard | Yes, handed over when the case reaches it | Not stated | Scorecard against the firm’s bar, with a verdict | Not stated | Not stated |
| CaseWithAI | Not stated; interviewer video with live transcript | Not stated for cases | Not stated | After every answer in coaching mode | Not stated | Not stated |
| CasePrepared | Yes, voice only | Charts to analyze; display not stated | Not stated | 4 dimensions, expected answers, filler words, think time | Not stated | Not stated |
| Case Study Prep AI | Yes | Not stated | Not stated | Structure, math, synthesis | Not stated | A case generated from a job posting |
| ChatGPT, Claude or Gemini | ChatGPT voice mode | No, the exhibit is read out | Only if you push it | Only against a key you paste in | No | You write the prompt |
Two columns in that table matter more than they look. Where the interview happens decides whether your practice includes the video call itself. Whether you can bring your own case decides whether your club’s casebook, or the case your target office is known for, can become practice at all.
How to choose
Start from what you are short on, not from the feature list. Most candidates are not short on cases to read. They are short on realistic reps, on someone to catch their math, and on time.
If exhibits and arithmetic under pressure are where you lose points, pick a tool that puts the chart on your screen and makes you find your own errors. If you have an online test coming as well, a platform with simulators for it saves you a second subscription. If you are interviewing with firms beyond McKinsey, BCG and Bain, firm-specific calibration matters more than library size. If budget is the constraint, a strict ChatGPT or Claude prompt plus a few pay-per-case sessions gets you surprisingly far.
Whatever you pick, the loop matters more than the tool. Run one case end to end without stopping. Read the whole report, not just the score. Redo the same case the next day, which is the step people skip and the one that tells you whether the fix stuck. Then one new case a day, mixing case types and styles. Book a handful of human mocks for the last two weeks, for presence and firm cadence. The MBB case interview guide walks through that loop and one exhibit in detail.
Start here
If your first round is a few weeks out and you have not run a full case out loud yet, do that before comparing anything else. Pick a medium case, block thirty five minutes, put pen and paper next to you and run it like it counts. It will feel messier than reading the casebook did. That is the point. Then read the report and run the same case again tomorrow.
On Tough Tongue, the 38 cases are in the MBB collection and on the case interview prep hub, from easy ones like Aquador and Fruit Drink Co. to Lux Rides and High Q Plastics when you want to be tested. If you run a consulting club or a career center and want your own casebook turned into interviewers, book 15 minutes with me and I will show you how the build works.
“Let me pull up Exhibit A. What is the minimum feasible price?”
Start the case38 interviewer-led MBB cases, graded against each casebook’s answer key, in the browser or on Google Meet and Zoom.
- Pick a case from the MBB collection, easy to hard.
- Run it end to end with pen and paper ready.
- Read the report, then redo the same case tomorrow.
