This is now one of the first paths into the top consulting firm, and it is a quietly dramatic moment for hiring everywhere. The firm that spent decades selling clients the case for deploying technology has quietly become one of its earliest large-scale adopters. Entry-level roles feed every practice at the firm, and that is exactly where they started testing candidates against a machine. Where an organization pilots an AI before anything else is a strong read on how confident it has become.
What the first cut actually is
The trick with this question is that it assumes there is one cut. There is not. Every hiring pipeline runs two different decisions. The first is the funnel: take eleven thousand applicants, find the few hundred worth a second look, and do it without a recruiter reading the same line twice by mistake. The second is the verdict: choose the person, knowing that a bad hire on a client-facing role costs far more than a missed one. The first is a filtering function. The second is a judgment call. AI is good at the first. It is not finished with the second.
That distinction is worth sitting with, because it reframes the whole debate. Nobody should be shocked that McKinsey is testing entry-level candidates against a machine. At that level there is no track record to weigh, only raw signal: how someone structures a thought, handles pressure, and communicates on the spot. The first cut has always existed to reduce noise at scale, and an algorithm does that without a bad morning, without fatigue, and without caring whose connections happen to look like yours.
What the machine cannot see
Yet the same logic cuts the other way, and hiring leaders who ignore it will regret it. A ten-minute conversation captures a performance, and a performance is not the same thing as a decade of judgment. The candidate who aced the interview is not always the consultant who will sit across from an angry client at midnight, holding together a team that is falling apart. Machines are also easy to game, and a candidate who learns the shape of the simulation can produce the exact right-sounding answers without gaining anything the job actually requires.
Then there is the promise that AI simply removes human bias, which deserves its own skepticism. The risk is not that a machine is cruel. It is that a machine automates a preference and dresses it in arithmetic. If the training scenarios quietly reward one style of thinking, one background, or one way of speaking, everyone outside that shape is penalized by default, and the bias comes with a receipt instead of a gut feeling. The audit is the whole ballgame: who tested the tool, on whose data, and what it would take to overturn the score it produced.
There is also a question no algorithm can answer, and it is a legal one. An automated hiring decision is now a decision in a growing number of jurisdictions. You can appeal a recruiter. It is much harder to appeal a score that no one can fully explain, and the candidate denied entry to a career over a ten-minute exchange deserves a path to a human who can simply say why.
Where the line should sit
All of which brings us to the honest version of the honest question, and the answer is yes, with a condition. AI is ready to make the first cut, and entry-level is the right place to try it. The first cut is a filtering function, and filtering is exactly what the technology does well, so this reads as a division of labor rather than a surrender of it. But two conditions have to hold, or the tool stops being an assistant and starts being an authority.
The first is that the machine makes the cut and the humans make the call. The AI narrows the pool, and the people who own the outcome decide who ends up in it. The moment you let the score stand in for the decision, you have outsourced the very thing you cannot put back. The second is that the first cut stays reversible. If the machine's score is the final word, you have quietly hired an algorithm. If it is just the first data point on a human's desk, you have handed that person a genuinely better tool.
The reason entry-level makes sense here is that the first cut and the final hire are not the same risk. Wrong at the top, you can lose a leader who shapes the firm. Wrong at the bottom, you lose a candidate and you rehire cheaply. Prudent companies test their AI where the downside is small and the volume is large, then reserve their human scrutiny for the decisions that actually stick.
The answer, plainly
So is AI ready to make the first cut? Yes. And the reason is that the first cut is not really the cut. It is a gate, not a verdict, and gates are exactly the kind of thing machines should own. The decision that matters, the one that decides who leads the deal, who stays for a decade, and who your clients learn to trust, is not made by a machine in ten minutes. It is made by the people the machine leaves standing.