Some North American candidates for business analyst roles will spend 10 minutes talking with an AI client avatar while human assessors watch
The next client a McKinsey hopeful pitches may not technically exist. The consulting firm plans to have some entry-level candidates hold a 10-minute conversation with an AI simulation of a client during their interviews, according to The Wall Street Journal.
The twist is that the software will not grade anyone. Humans stay in the room and decide how each candidate performed. McKinsey has handed AI the role of the client, not the role of the judge.
The pilot targets undergraduate candidates applying for entry-level business analyst roles. Those jobs come with a starting salary of $128,000, so the stakes for applicants are hardly trivial.
The program starts this fall in North America. It is currently part of the recruiting process for hundreds of candidates in the region, not the firm’s entire global applicant pool.
Each candidate gets roughly 10 minutes with an AI avatar built to play a client. The exercise is meant to mirror an ordinary conversation between a consultant and the business paying for advice.
McKinsey partner Saleem Zafar said the avatar covers approximately 80% of the interactions consultants usually have.
What the bot will not do is lose its temper. According to the WSJ, the simulation is not designed to “mimic an unhinged client.” Candidates can expect a reasonable executive, not a scene from Succession .
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
This is not McKinsey’s first experiment with AI in recruiting. The firm brought Lilli, its generative AI tool, into final-round interview stages starting in January 2026.
In April 2026, McKinsey launched a free AI practice tool. It helps candidates prepare for the quantitative parts of case studies, the math-heavy portion that tends to trip people up under pressure.
The client avatar sits alongside the firm’s existing hiring steps rather than replacing them. Those include personal experience interviews and the Solve assessment, a game-like test that already made the process feel less like a traditional job interview.
The human-scoring design addresses an obvious worry. Candidates often fear being rejected by an algorithm they cannot see or question. Keeping assessors in the room means a person, not a model, is responsible for every judgment call.
It also gives McKinsey a cleaner comparison. When every candidate faces the same simulated client, assessors can watch how different people respond to similar prompts. A human interviewer playing the client might vary from session to session, while software can hold the script steady.
The open question is scale. Right now the program covers hundreds of candidates in a single region.
For now, the arrangement is modest by design. The AI plays the client for about 10 minutes, the humans keep the scorecard, and the $128,000 offer still depends on impressing real people.