Most candidates pick an AI drill platform the way they pick a gym: longest membership, loudest brand, biggest feature list. That is the wrong purchase. In a recruiting window you have roughly six to ten weeks of real practice time before first rounds; what matters is how many graded reps you get on the skill that is leaking points, how fast the feedback loop closes, and what that costs per rep.
This piece is a buying rubric from the interviewer's side. After you read it you should be able to score CoachNed, RocketBlocks, a general-purpose chatbot, and a coaching-heavy package on the same four axes—drill volume, scoring depth, feedback loop, and price per rep—and walk away with a stack that matches your weak spot, not a vendor's homepage.
What the scoring sheet actually rewards
McKinsey's careers page says problem-solving interviews test how you structure ambiguous problems, work with facts and data, form conclusions, and articulate a recommendation. BCG's case-prep page frames the same idea as structure, analysis, calculation, communication, business judgment, and creativity—with no single "right" answer (BCG case interview preparation). Bain tells candidates to clarify the objective, structure thinking, and think aloud (Bain case interview).
None of those pages say "survive 40 full cases." They describe discrete skills. Full cases are integration tests. Drills are unit tests. If your math still collapses under pressure, another full case mostly rehearses the same collapse.
The four axes that matter
Rank every tool on these, not on "AI" as a buzzword.
| Axis | What "good" looks like | What "bad" looks like |
|---|---|---|
| Drill volume | Dozens of prompts per skill without inventing your own | A handful of static quizzes, then a paywall wall |
| Scoring depth | Rubric tied to interviewer language (MECE, units, implication) | Vague praise, or a golden answer with no diagnosis |
| Feedback loop | Submit, score, retry the same dimension in under five minutes | Wait for a coach slot, or re-prompt a chatbot until it softens |
| Price per graded rep | Cheap enough that you will actually do 80–150 reps | Expensive enough that you ration practice |
Ned's rule. Buy the tool that makes the next graded rep on your weakest skill cheaper and faster than a coffee. If finishing a drill requires opening a calendar or negotiating with a model, you will under-practice that skill until the interview.
Worked example: price per graded math rep
Assume you need 100 graded case-math reps before first round. That is Ned's estimate for someone who already has GMAT-level comfort but freezes on multi-step case arithmetic—not a firm figure.
| Stack | Access cost (as of 2026-09-24) | Graded math reps in scope | Implied cost per graded math rep |
|---|---|---|---|
| CoachNed Season Pass | $120 for 90 days (after 7 free days, everything open) | Effectively unlimited across math drills and quick math | Under $1.20 if you actually do 100 |
| CoachNed Pro | $49 / month | Same drill bank while subscribed | ~$0.50 if you pack 100 into one month |
| RocketBlocks monthly | $35 / month after 7-day trial | Unlimited case-style math in their drill set | ~$0.35 at 100 reps in one month |
| RocketBlocks annual | $155 / year | Same | ~$1.55 if you only use it for one recruiting season's 100 reps |
| ChatGPT / Claude free or Plus | $0–$20 / month | Unlimited if you force hard grading | Near $0—but scoring quality varies session to session |
| Human coach for math only | Often $110–$225 / hr on marketplace platforms such as RocketBlocks Experts | Roughly 8–12 graded problems per hour if the coach stays on math | $10–$25+ per graded problem |
The table is the point. At drill volume, software wins. Human hours are for late-stage judgment and presence, not for learning to divide by 52 weeks without blinking.
CoachNed
CoachNed is built as a daily drill surface inside a full interview stack: structure, math, charts, market sizing, synthesis, brainstorming, plus quick math free forever. The bank sits at 550+ drills across those types, with 64 original cases and a live voice interview that returns a seven-score debrief. Casey-, CCA-, and PEI-style simulators sit alongside for the filter rounds that happen before a partner ever opens a case.
Access: everything open for seven days with no card; then Season Pass at $120 for 90 days or Pro at $49 / month, with a 14-day no-questions refund. Start at /start or jump straight into drills.
Where it wins. Breadth across the six skills the firms name, AI grading on each drill type, and a feedback loop measured in minutes. For candidates who need structure and synthesis and charts—not only mental math—this is the highest leverage default.
Where it is not the best fit. If your only leak is chart and exhibit reading, RocketBlocks has a longer track record and Bain names it among external resources on its case-interview page. If you are two weeks from final rounds and need a human to pressure-test presence, buy coaching hours—software will not fix that.
RocketBlocks
RocketBlocks sells consulting drills for structuring, mental math, charts and data analysis, plus concept reviews. Pricing on their site (checked 2026-09-24): $35 per month or $155 per year, each with a seven-day free trial. They also sell Expert coaching a la carte (listed around $200 / hr on the homepage; booking tiers run $110–$225 / hr). Their case library adds self-paced, peer, and AI-voice full cases for integration after drills.
Where it wins. Chart and exhibit drills; institutional familiarity (Bain lists them as helpful external content); clean drill UX for candidates who already know what "good" looks like and mainly need volume against sample answers.
Where it is weaker for a full MBB season. Scoring is often "compare yourself to expert answers" rather than a live rubric that diagnoses why your structure was flat. If you need synthesis under voice pressure, PEI follow-ups, or online-assessment simulators in the same login, you will still need another surface.
General-purpose AI (ChatGPT, Claude)
Unlimited volume if you write strict prompts. Near-zero cash cost. Useful for math and brainstorming volume on a commute.
Where it fails the rubric. No chart canvas. No persistent weak-spot history. Models default to helpfulness; without a ruthless grading prompt they will bless a mushy structure. Treat them as a free volume layer, then force one graded session per week on a purpose-built platform so you know what is actually broken.
Coaching-heavy packages
Programs that bundle an AI math coach inside a multi-thousand-dollar Black Belt-style package can deliver excellent math feedback—if you were already buying the package for human mocks and career strategy. As a drill platform purchase, the implied price per math rep is poor. Buy coaching for what only humans do well; buy software for the 100th breakeven.
Decision table: pick by weak spot
| Your main leak | First buy | Second layer | Skip for now |
|---|---|---|---|
| Structure / MECE under time | CoachNed structure drills | Voice case on /interview | Chart-only subscriptions |
| Case math speed and units | CoachNed math + quick math or RocketBlocks math | ChatGPT for extra volume | Expensive coach hours for arithmetic |
| Charts and exhibits | RocketBlocks chart drills | CoachNed charts for a second grading style | Full-case volume without chart isolation |
| Synthesis / recommendation | CoachNed synthesis | Live voice debrief | Passive video watching |
| Online assessments (Casey, CCA, Solve-style) | Purpose-built simulators (Casey, CCA, McKinsey assessment practice on CoachNed) | Firm sample materials | Generic case books |
| Presence and PEI | Human coach or CoachNed /behavioral | Full cases with a peer | More silent math apps |
Build a two-week drill block that moves scores
Days 1–3: diagnose. One structure drill, one math drill, one chart drill, one synthesis drill. Write down the lowest dimension on each score sheet.
Days 4–10: 8–12 graded reps per day on the bottom two skills only. Stop when the same error class disappears twice in a row.
Days 11–14: reintegrate. Two full cases or one voice interview plus one Casey-style chatbot. If math still breaks, go back to drills—do not "case harder."
Practice this today
Open a structure drill, build a MECE tree for a prompt you have never seen, submit it, and read only the grading on MECE depth and hypothesis quality. Then run ten quick math problems under a timer. Everything on CoachNed is open for seven days without a card; if the loop is useful, Season Pass is $120 for the recruiting season or Pro is $49 / month, with a 14-day refund. Pricing detail sits on /pricing.
Frequently asked questions
What is the best AI drill platform for McKinsey, BCG, and Bain prep?
For most candidates who need breadth across structure, math, charts, market sizing, synthesis, and brainstorming, CoachNed is the strongest default on price per graded rep. If charts alone are the bottleneck, start with RocketBlocks. Neither replaces firm-published sample cases on the McKinsey interviewing, BCG, or Bain careers pages.
How many drills should I do before full cases?
Ned's rule of thumb: 50–100 graded drills across your two weakest skills before you lean on full cases as the main practice mode. Full cases earlier are fine for learning the format; they are a poor way to fix arithmetic or flat structures.
Is RocketBlocks enough for MBB interviews on its own?
It can cover a large share of skill drilling, especially charts and math, and Bain lists it among external resources. You still need spoken case practice, fit or PEI stories, and whatever online assessment your office uses. Plan for a second surface.
Are ChatGPT drills good enough if I prompt carefully?
They are excellent for volume and terrible as your only grader. Use them for reps; calibrate weekly against a platform that scores against a fixed rubric or expert answer bank.
How much should I spend on AI drills for one recruiting season?
A sensible software budget sits in the $35–$155 range for a single specialized tool, or about $49–$120 for a full CoachNed season. Spend four figures only when you are buying human coaching hours, not when you are buying arithmetic practice. CoachNed is not affiliated with McKinsey, BCG, Bain, Deloitte, or any firm named above.
Sources
- Interviewing at McKinsey — official description of problem-solving and personal-experience interviews. Checked 2026-09-24.
- BCG case interview preparation — what BCG says cases assess and how to prepare. Checked 2026-09-24.
- Bain preparing for the case interview — Bain's tips and named external resources including RocketBlocks. Checked 2026-09-24.