Why McKinsey is testing candidates with an AI client

McKinsey is piloting an AI client in its interviews: a high-stakes conversation in a low-stakes setting. Here’s how it works and how to run your own.

McKinsey is piloting an AI client in its interviews: a high-stakes conversation in a low-stakes setting. Here’s how it works and how to run your own.

McKinsey is piloting an AI client in its interviews: a high-stakes conversation in a low-stakes setting. Here’s how it works and how to run your own.

· 5 min read

Collage of a hand adjusting speech bubbles across from a geometric paper client figure

CATEGORY

Assessments

PUBLISHED

READING TIME

5 min

The short answer

This fall, McKinsey began piloting a new step in some entry-level interviews: a short spoken conversation with an AI client. The candidate gives a recommendation, and the AI, playing someone like a city mayor, raises new concerns and adds context as the conversation goes. Partners in the room judge how it went. The AI gives no score.

The technology is the headline, but the format is the idea worth borrowing: a high-stakes client moment, in a setting where nothing real is on the line, run the same way for every candidate.

What McKinsey is doing

The Wall Street Journal reported on October 14, 2026 that McKinsey is adding “a new type of client simulation” to its interviews, to test how recruits think on their feet and handle client conversations. Here’s what has been reported so far.

  • Who: undergraduate Business Analyst candidates in some North American offices. Hundreds are expected this fall, and other roles may follow in 2027.

  • Format: about 10 minutes, on a laptop in a McKinsey office, after the case interview. The case and personal experience interviews stay.

  • Scenario: in one example from Dallas partner Saleem Zafar, a city hires McKinsey to pick a site for a new stadium, and the AI plays the mayor.

  • Personas: they include a “professorial” client who wants the reasoning behind a recommendation. The AI isn’t built to be unhinged or to ask gotcha questions.

  • Scoring: people do it. McKinsey’s careers site says “AI is not used to make autonomous hiring or interview-selection decisions.”

It follows a spring 2026 pilot in which some final-round candidates had a 20-minute “client conversation” with a fictional stakeholder. McKinsey told candidates that version wasn’t evaluated.

Why role play beats another interview question

Most interviews ask candidates to describe a hard moment from their past. That shows you who prepared a good story. Role play shows you how they handle one as it happens.

Zafar described the goal this way: “What we’re looking for is how do you respond to changing stimuli?” In practice that comes down to three questions. Did the candidate hear what mattered to the client? Did they act on a new concern when it came up? Or did they push through with the answer they walked in with?

You can rehearse an answer. You can’t rehearse a client who changes the brief.

Why an AI plays the client

Firms have used human role players for years. The trouble is consistency. As reported by the Times of India, McKinsey partners point out that some human interviewers play tougher clients than others. An AI keeps the persona the same, so every candidate meets the same mayor, with the same priorities, raising the same concern at the same point.

That consistency is what makes an interview worth trusting. A 2022 meta-analysis by Sackett and colleagues found structured interviews among the strongest predictors of job performance, with an estimated validity of .42, against .19 for unstructured interviews. A role play where every candidate faces the same persona and the same twist is a structured interview, just a more realistic one.

McKinsey reportedly designed its AI client to reflect about 80% of the situations consultants meet day to day. That figure is a partner’s estimate, not a measured result.

What it means outside consulting

Most jobs have a version of the mayor. An account executive meets a buyer whose budget was just cut. A customer success manager hears from a customer who is ready to cancel. A recruiter works with a hiring manager who changes the brief halfway through. These moments decide how someone does in the role, and a standard interview rarely shows them.

McKinsey can run role play late in the process because partners sit in on each session. Most teams don’t have a senior person free for every candidate, so role play gets saved for the final round or skipped altogether.

How to run a role play interview

  1. Pick the moment the job is made of. Write the persona: who they are, what they want and what worries them.

  2. Plan one twist. Partway through, the persona raises a new concern or changes a fact. Every candidate gets it at the same point.

  3. Decide what good looks like before anyone starts. For example: did they restate the concern, did they adapt or explain why the plan still holds, and was the recommendation clear?

  4. Keep it short. McKinsey’s version runs about 10 minutes.

  5. Keep people on the decision. Use the role play as evidence for your team, not as an automatic cut.

How TailorHire runs role play

TailorHire’s Role Play brings this format to the first round. You describe the persona and the scenario, TailorHire’s interviewer plays the part live, and every candidate gets the same scenario and the same twist. Each criterion is rated out of 5 and linked to the moment in the conversation behind it. Your team makes the call, and no one is declined automatically.

See how Role Play works →

McKinsey hasn’t said what technology powers its pilot, and TailorHire isn’t affiliated with McKinsey.

FAQ

Does McKinsey’s AI decide who gets hired?

No. McKinsey says AI is not used to make autonomous hiring or interview-selection decisions. Partners and assessors judge the conversation.

Who takes McKinsey’s AI client interview?

As reported, undergraduate Business Analyst candidates in some North American offices, starting in fall 2026.

Can candidates prepare for it?

McKinsey says no additional preparation is required. Prep coaches suggest practicing recommendations that get interrupted and restating the client’s concern before answering.

Is role play fair to candidates?

It’s fairer when it’s consistent. The same persona, the same twist and the same criteria for everyone mean candidates are compared on how they handled the same moment.

Sources: The Wall Street Journal (Oct 14, 2026), as reported by The Verge, the New York Post and the Times of India · McKinsey, AI at McKinsey (careers page) · Management Consulted, McKinsey client conversation pilot (May 2026) · Sackett et al., Journal of Applied Psychology (2022)

See a two-way interview run.

Book a demo

See a two-way interview run.

Book a demo

SHARE

Find your next

hire.

Bring one open role. We’ll set up the first screens with you.

Find your next

hire.

Bring one open role. We’ll set up the first screens with you.

Find your next

hire.

Bring one open role. We’ll set up the first screens with you.