The McKinsey AI Interview Tests Listening

Summary: In the new McKinsey AI interview, some entry-level candidates face a ten-minute conversation with an AI “client”, and the AI will not score anyone. McKinsey is using AI to test the human part of consulting, and the logic applies to partner hiring.

The McKinsey AI interview puts an avatar in the client’s chair and keeps it out of the hiring decision.

According to the Wall Street Journal’s report, some Business Analyst candidates in North America now face a ten-minute conversation with an AI avatar after their case interview. It plays the client, questioning the recommendation and adding context.

Partners watch. The AI gives no score. Hundreds of candidates will take part this autumn. Other roles may follow next year.

In the firm’s example, the avatar plays a mayor deciding where to build a stadium, raising new concerns as the candidate presents.

This is McKinsey’s second AI change to hiring this year. In January, Fortune reported a final-round pilot in which candidates work with Lilli, McKinsey’s internal AI tool, and are judged on how they use it. The first test asked whether candidates can work with AI. This one asks what they can do that AI cannot.

The most revealing part of McKinsey’s new interview is what the technology will not do: decide who gets hired.

What the McKinsey AI interview is really testing

The McKinsey AI interview tests whether a candidate can listen and adapt when a client changes the question, not whether they can produce an answer faster than a machine.

Saleem Zafar, the Dallas partner who helped design the pilot, told the WSJ he is watching for whether candidates heard what mattered and acted on it. The failure he describes is the candidate who arrives with their own agenda and steamrolls the answer.

That is a test of conversation, not analysis.

The case interview rewards structured problem-solving. This simulation sits closer to the real job: defending a recommendation while the client moves the goalposts. Most consultants know the mayor who agrees with the logic and then raises a political problem nobody modelled.

It fits a wider shift. The firm has been hiring for judgement over pedigree, and CEO Bob Sternfels told HBR’s IdeaCast that McKinsey is “looking more at liberal arts majors, whom we had deprioritized”.

AI is being used to make human skills visible

McKinsey is using AI to make distinctly human qualities easier to see, not to test technical AI capability.

That is the irony. The avatar is not judging anyone’s skill with a model. It creates a setting where listening and composure under pressure become visible.

The firm’s own build-out explains why. Bob Sternfels told HBR’s IdeaCast that McKinsey went from roughly 3,000 AI agents to about 20,000 in around 18 months, alongside a workforce of about 40,000 people.

When agents do more of the analysis, it stops separating candidates. What separates them is how they frame the problem, and whether a client would trust them with the answer.

Junior consulting work is changing first

Junior consulting work is shifting from producing analysis to handling clients, and McKinsey’s hiring criteria are moving ahead of that change.

If AI performs more of the research and modelling, firms need to spot communication and client judgement at the point of hire, not three years in. We have argued that AI is killing the old hiring model rather than consulting itself. This is what that looks like inside one firm.

It also reshapes the consulting apprenticeship model. Juniors learnt judgement by grinding through the analysis. Remove much of the grind and firms must either teach judgement differently or hire people who already show it.

McKinsey appears to be doing the second.

The McKinsey AI interview targets the coaching industry

The McKinsey AI interview is partly a response to the interview-coaching industry.

Blair Ciesil, the partner who co-leads McKinsey’s global recruiting, told the WSJ that candidates “can come in with stories that are quite rehearsed”, which makes it hard for interviewers to reach the “underlying intrinsics”. That is a candid admission. The case interview, consulting’s signature filter, has been prepped to death.

The avatar adds movement. It is built to reflect 80% of the scenarios McKinsey consultants meet in daily work, with personalities such as the “professorial” client who wants to understand the reasoning. Its value is that the conversation shifts in ways a script cannot fully anticipate.

It will not end preparation, only change it. If the format sticks, coaching providers will quickly build their own avatars. McKinsey may simply have started a new preparation cycle.

Standardisation is not the same as objectivity

A standard AI client makes the interview stimulus consistent, but human assessors still decide what good looks like.

McKinsey partners describe the tool as adding fairness, because some human interviewers play the client harder than others. That is a genuine improvement.

It does not remove judgement. Assessors still decide what counts as good listening, appropriate confidence and real adaptability. Those are the qualities where bias tends to hide: in which style reads as confident, and whose pushback reads as composure rather than defensiveness.

Consistency of input is not objectivity of outcome. Firms that copy the format should be clear about which of the two they are buying.

Realism does not prove predictive validity

A more realistic interview exercise is not automatically a more predictive one.

Saleem Zafar framed McKinsey’s goal plainly: predicting whether a person will be successful. That claim needs evidence. McKinsey will need to show that performance with an AI mayor predicts performance with a real, politically sensitive client, one with a board behind them and a reputation at stake.

McKinsey hires at scale, so it can test this. If the data holds, the format will spread across the industry. If it does not, it becomes well-produced theatre.

Partner hiring has a bigger rehearsal problem

The most rehearsed interview in consulting is not the graduate case. It is the partner business case.

A lateral partner arrives with a polished story of their book of business and a three-year plan, refined over several rounds with several firms. Knowing what it takes to make partner and pitching it well are different skills.

Yet lateral partner hiring rarely tests what happens when the client changes the question. Firms read the plan, check references and run chemistry meetings. Few put the candidate in a room with a sceptical CFO who keeps moving the goalposts.

The steamroller habit McKinsey is screening out at analyst level is the one clients still complain about at partner level. A senior hire who arrives with the answer before hearing the problem is expensive, and usually discovered after the first client meeting.

The case was never the real test. The client conversation was.Ben Appleton, Strat-Bridge

As AI makes senior consultants more valuable, the cost of a wrong partner hire rises with it. This is where we spend most of our time with firms hiring senior consulting leaders.

What the McKinsey AI interview means for hiring leaders

The firms that gain most will treat the McKinsey AI interview as a prompt, not a template. They will test how candidates handle a client who changes the question at every level, partners included, and they will measure whether it predicts anything.

What they will still get wrong is assuming a new format solves rehearsal. It moves it. The candidates who prepare hardest will learn the avatar, just as they learnt the case. The edge will sit with firms that keep changing the stimulus and keep a human accountable for the call.

Would your last partner hire have survived ten minutes with the mayor?

Ben Appleton

Ben Appleton is the founder of Strat-Bridge, a specialist executive search partner to the management consulting industry. He works with global consulting firms and senior leaders across the UK, Germany, Switzerland, and beyond — helping them build capability at the Partner and Director level.
Connect with Ben on LinkedIn or email ben@strat-bridge.com.