McKinsey Tests an AI Client Simulation for Some Job Candidates

McKinsey & Company is testing a new interview exercise in which some entry-level candidates speak with an artificial-intelligence simulation of a client. The trial is aimed at undergraduates applying for business-analyst roles in some North American offices during fall 2026.

The exercise reflects a broader shift in recruiting: employers are beginning to use AI not only to screen applications, but also to recreate workplace situations and observe how candidates respond. In McKinsey’s case, however, the AI is not intended to make the hiring decision. Human interviewers will assess the candidate’s performance.

How the McKinsey AI interview works

According to reporting by The Wall Street Journal, some candidates will have a 10-minute conversation with a simulated client. The system is designed to represent roughly 80% of the situations McKinsey consultants may encounter in their ordinary work, according to the report.

The simulation is not meant to behave like an extreme or deliberately hostile customer. The tool is not designed to “mimic an unhinged client.” That distinction suggests the exercise is focused on routine consulting skills rather than a stress test built around confrontation.

Candidates may need to:

• Clarify an unclear business problem
• Ask focused follow-up questions
• Explain an idea in accessible language
• Respond when the client changes direction
• Organize information under time pressure
• Maintain a professional, collaborative tone

The AI will not independently score applicants. Human interviewers will be present to evaluate how candidates handle the conversation and the situations presented by the simulation.

The trial may involve hundreds of candidates, although the source material does not establish the precise number or identify every participating office.

What McKinsey is actually testing

A conventional case interview often gives candidates a structured business problem and asks them to reason through it aloud. A simulated client conversation tests a somewhat different set of abilities.

The candidate is likely being judged less on producing a perfect answer than on the quality of the interaction. Strong performance could involve:

1. Defining the problem before solving it
   A candidate who immediately proposes solutions may miss the client’s real concern. Asking what outcome the client wants can establish a clearer direction.

2. Listening for constraints
   Budget, timing, organizational resistance, data quality, and implementation capacity can matter as much as the headline problem.

3. Structuring the discussion
   Even a short conversation benefits from a clear sequence: understand the issue, identify the major drivers, test assumptions, and agree on next steps.

4. Adapting without losing focus
   A simulated client may introduce new information or challenge an assumption. The candidate needs to acknowledge the change while keeping the conversation coherent.

5. Communicating with judgment
   Consulting requires more than analytical ability. Candidates must explain uncertainty without sounding evasive and make recommendations without overstating what the evidence shows.

The format may therefore reward candidates who combine analytical thinking with the interpersonal skills traditionally assessed through case interviews, behavioral questions, and role-playing exercises.

Why use an AI client?

An AI simulation can give interviewers a more consistent way to present a scenario to many candidates. A human interviewer may unintentionally vary the wording, level of prompting, or degree of difficulty from one conversation to another. A standardized system could reduce some of that variation.

The format may also allow a firm to assess behavior that is difficult to capture through a résumé or a conventional question such as “Tell me about a time you solved a problem.” Speaking with a simulated client can reveal whether an applicant:

• Asks questions before making assumptions
• Notices ambiguity
• Handles incomplete information
• Explains complex points clearly
• Remains composed when challenged

That does not make the exercise a complete measure of consulting ability. A 10-minute interaction captures only a narrow slice of a candidate’s skills, and performance may be affected by familiarity with conversational AI, accessibility needs, language background, or discomfort with an unfamiliar format.

The AI may create the scenario, but people remain responsible for interpreting the candidate’s response.

The risks of AI-based interview exercises

Using AI in recruitment raises concerns even when the system is not making the final decision.

Consistency is not the same as fairness

A standardized scenario can make comparisons easier, but candidates do not all experience the technology in the same way. Some may be more comfortable speaking to an artificial interlocutor than others. Differences in accent recognition, speech patterns, disability access, or technical performance could influence the interaction.

The source material does not provide details about the system’s accessibility safeguards, how technical problems will be handled, or whether candidates can request an alternative format. Those policies will affect how fair the exercise is in practice.

Human judgment still requires discipline

Removing automated scoring does not remove subjectivity. Human interviewers will still decide what counts as a strong response and how much weight to give the exercise.

A candidate who speaks confidently may appear more capable than someone who reasons carefully but communicates less smoothly. Clear evaluation criteria can help interviewers distinguish polished presentation from genuine listening, judgment, and problem-solving.

The simulation cannot fully reproduce a client relationship

Real consulting conversations develop over time. Clients may have political concerns, competing priorities, incomplete knowledge, or established relationships with the consulting team. A short artificial exchange can approximate some communication patterns, but it cannot reproduce the full context of client work.

The exercise should therefore be treated as one input into a broader assessment rather than a definitive test of whether someone can succeed as a consultant.

How candidates can prepare

Candidates do not need to predict the exact wording of the AI simulation. Preparation should focus on habits that transfer across many client conversations.

Start with the objective

Before offering an answer, establish what the client is trying to achieve. Useful questions might include:

• What outcome would make this project successful?
• What has changed to make the issue urgent?
• Which constraints cannot be compromised?
• Who will be affected by the decision?
• What information is already available?

The goal is not to ask questions for their own sake. Each question should reduce uncertainty or help determine what to do next.

Use a simple structure

A concise framework can keep a short conversation from becoming scattered:

1. Restate the problem in your own words.
2. Confirm the desired outcome.
3. Separate the issue into two or three relevant areas.
4. Identify the information needed to assess those areas.
5. Summarize the implication and propose a next step.

This approach is more useful than reciting a memorized consulting framework that does not fit the situation.

Make assumptions visible

If information is missing, say what assumption you are making and how it affects the recommendation. For example, a candidate might explain that a proposed course of action depends on the organization having enough capacity to implement it.

Visible assumptions give the interviewer a way to evaluate the candidate’s reasoning. They also reduce the risk of presenting speculation as fact.

Respond constructively to new information

If the client introduces a constraint, do not ignore it or defend the original answer reflexively. A strong response might acknowledge the change, explain which part of the analysis it affects, and adjust the recommendation.

That demonstrates flexibility without abandoning structure.

End with a clear summary

A short client conversation should conclude with the candidate making the current position explicit:

• What the central issue appears to be
• What remains uncertain
• What action should come next
• What evidence would help confirm the recommendation

A clear summary can be more valuable than trying to cover every possible angle.

What this means for the future of recruiting

McKinsey’s trial is part of a wider movement toward AI avatars and simulated interview experiences. Other companies are also developing tools intended to recreate aspects of a job interview or workplace interaction.

The appeal is straightforward: simulations can expose candidates to situations that are difficult to test through written applications. They may also help employers run more consistent exercises across applicants.

But the quality of the process will depend on details that are not yet clear from the reported trial:

• How candidates are informed about the system
• Whether conversations are recorded
• How long any recordings or transcripts are retained
• What accommodations are available
• How interviewers are trained to assess the results
• Whether candidates can challenge a technical or procedural problem
• How much weight the exercise carries in the final decision

Those questions determine whether AI is being used as a useful simulation or as an opaque additional barrier in an already competitive hiring process.

The bottom line for applicants

McKinsey’s AI client exercise appears to be a human-evaluated role-play, not an automated interview in which a machine assigns candidates a score. For applicants, the most practical preparation is to practice concise, structured conversations: clarify the objective, ask purposeful questions, explain assumptions, adapt to new information, and finish with a recommendation or next step.

The experiment will also test McKinsey’s process. If the firm can use the simulation consistently while preserving accessibility and meaningful human judgment, it may become a useful complement to case interviews. If the technology introduces new sources of bias or measures comfort with AI more than consulting ability, its value will be considerably narrower.

Sources

• The Verge, “McKinsey is making job applicants spar with a fake AI client,” October 14, 2026: https://www.theverge.com/news/1011654/mckinsey-ai-client-interview