You Can't Watch Your Way Into an Offer

I spent five years interviewing candidates at BCG. The ones who failed usually weren't underprepared. They were overprepared in the wrong way.

They'd watched every case video on YouTube. Read the casebooks. Memorized the frameworks. Then they'd sit down across from me and dive into a problem with no structure, force-fit a framework that didn't match, or nail the math and completely miss what it meant for the client. Forty cases of preparation, and the same mistakes a first-timer makes, just delivered more smoothly.

Here's the uncomfortable truth I learned on the other side of the table: most interview prep is watching. And watching doesn't build skill.

Employers are done taking your word for it

For a long time, a strong resume and a good school were enough of a signal. That's over. AI wrote the resumes, so employers stopped trusting them. In one 2025 survey of over 2,000 employers and job seekers, most employers said they can no longer verify skills from a resume at all, and more than half have dropped degree requirements entirely. What's replacing credentials? Demonstration. Eighty-five percent now use some form of skills-based hiring. Google brought back in-person interviews partly to see candidates think without an AI whispering in their ear.

And at the top of the market, the signal is unmistakable: McKinsey is piloting an AI-assisted final-round interview, where candidates work a case alongside the firm's AI while interviewers watch how they question it, judge its output, and apply the answer. The firm even released its own free AI practice tool because, in the words of its talent lead, doing the work is one thing, "doing it while somebody's watching you is something else."

The case interview isn't dying. Its logic (show me how you think, live) is spreading to everyone. Increasingly, that includes how you think with AI.

What interviewers actually score

Here's what nobody tells you: interviewers aren't grading your frameworks. At BCG, I scored hundreds of candidates on a rubric that came down to five things.

Structure. Do you organize the problem before attacking it, or dive in and hope? Strong candidates take thirty seconds to lay out a plan. Weak ones start calculating.

Hypotheses and logic. Do you form a point of view early and test it with sound reasoning as new information arrives? Or do you just collect facts and wait for the answer to appear?

Driving the case. Stating the what isn't enough. After each number or exhibit, candidates need to be able to drive the case forward with the "so what" and "now what."

Numbers in conversation. Can you do the math while talking, and then say what the number means? A right answer with no "so what" scores lower than people think.

Synthesis. When I say "the CEO walks in, what do we tell her?", can you answer the client's question in three sentences? Not a recap of your work. An answer.

Notice what these have in common: they're behaviors, not knowledge. You can't learn them by reading about them, any more than you can learn a golf swing from a book. And because they're behaviors, they're observable, coachable, and measurable.

Passive prep feels productive. That's the trap.

Watching case videos is the single most popular way students prepare, and it's the least effective. This isn't just my opinion from the interviewer's chair. It's one of the most robust findings in learning science. A landmark analysis of 225 studies in PNAS found students in passive, lecture-style learning were 1.5 times more likely to fail than students who practiced actively. Decades of research on expert performance says the same thing: skills are built through practice at the edge of your ability, with immediate feedback on specifics, not through exposure.

Watching someone else case is like watching someone else lift weights.

To be fair, students know this, which is why they do peer mocks. Practice with a partner is genuinely better than videos, but your practice partner is usually another 21-year-old who has also never sat on the other side of the table. You're getting reps without calibration. The alternative is a coach who does know what interviewers score, at $200 to $500 an hour. Real feedback has been a luxury good.

"Case ready" should be a measurement, not a feeling

This is the part that drove us to build CaseReady. Every practice session on our platform is a live, AI-run case or behavioral interview, scored against the same rubrics we used as interviewers at BCG: structure, hypotheses and logic, numbers, synthesis. You get a baseline, a trend line, and the specific skill to work on next. The loop that used to cost $500 an hour, on demand.

We're early, and we're honest about that. We're learning from every session students complete. But we believe the burden of proof is coming for everyone in this industry. Even the professors pioneering AI simulations in business schools admit they can't yet measure whether their methods work. Prep that can't show improvement is just content.

So here's the standard we'd hold anyone to, ourselves included: if your preparation isn't giving you a score, a trend, and a next step, you're not training. You're spectating.

The interview has changed. The resume is no longer proof, the frameworks are free, and the firms themselves are putting AI in the room. The only thing left that can't be faked is the skill itself.

Go build it. Deliberately.