AI in Job Interviews: How Candidate Assessment Is Changing in 2026

• #Assessment

Until recently, using ChatGPT during a job interview was generally considered a violation of the rules. In 2026, some international employers are taking a different approach: candidates are explicitly permitted to use AI, and their ability to apply the technology effectively to solve a business problem is assessed.

McKinsey is a notable example. According to the Financial Times, the firm has introduced an AI-enabled exercise into some interviews for entry-level consulting roles. Candidates work through a client case, using a tool of their choice to conduct research, analyse data or test ideas, and then explain which insights they drew from the model's output and why.

This does not mean that independent assessment is being abandoned. On the contrary, the new format helps distinguish a candidate's own reasoning from their ability to enhance their performance with technology.

The key trend can be summarised as follows: companies are gradually shifting from asking, "Has the candidate used AI?" to asking, "How does the candidate use AI, validate its output and make the final decision?"

From Prohibition to Controlled Use

The assessment landscape is evolving in two directions. Some employers are strengthening proctoring measures, conducting in-person assessments and prohibiting external assistance to evaluate candidates' knowledge without technological support. Others are introducing AI-enabled assessments, in which the use of artificial intelligence is explicitly permitted and built into the assessment scenario.

There is no contradiction between these approaches. Employers need answers to two questions:

• What can the individual do independently?
• How effectively can they work in an environment where AI has become a routine tool?

As a result, a hybrid format is emerging as a promising approach: candidates complete part of the exercise without AI, are then given access to the tool, and must subsequently defend their final recommendations independently.

How an AI-Enabled Interview Works

The new exercise follows the same principles as a traditional case interview. Candidates first familiarise themselves with a business problem, formulate hypotheses and analyse the available data. At a certain stage, the interviewer invites them to use AI to broaden the range of ideas or test a specific part of their proposed solution. The candidate then develops and presents the final recommendation without relying on the model.

For example, candidates may be asked to assess a new product launch, a change to the operating model or an investment in technology. They can use AI to identify additional risks, explore segmentation options or develop alternative scenarios. However, candidates must determine which information is genuinely useful, identify unsupported assumptions made by the model and establish which conclusions are substantiated by the underlying data.

The assessment does not focus on the number of tools used or the complexity of the prompt. In an interview with McKinsey's Director of Talent Assessment, the emphasis is on the fact that technical techniques can be taught. Critical thinking, intellectual curiosity, clear communication and the ability to challenge a confidently presented but flawed answer are far more important.

It is also important to note that permission to use AI does not apply automatically throughout the entire recruitment process. McKinsey's careers website states that candidates may use AI during assessments only when explicitly permitted by the organisers.

What Is Being Assessed?

In a traditional interview, a good final answer could reflect a lucky guess or a rehearsed response. In the new format, the final output alone becomes even less informative: a well-structured recommendation can now be produced quickly by almost any candidate with access to AI.

As a result, the focus is shifting towards observable behaviours:

• How the candidate structures the problem before turning to AI
• Whether they understand which parts of the task genuinely require the tool
• Whether they provide sufficient context, define criteria and set constraints
• Whether they validate figures, facts and reasoning
• Whether they identify contradictions and AI hallucinations
• Whether they can reject a model's recommendation
• Whether they can explain their reasoning
• Whether they take ownership of the final decision

The ability to recognise when not to use AI is particularly important. If a candidate turns to the model at every step, this may indicate not strong digital literacy but a lack of independent judgement and insufficient professional expertise.

Four Changes in Talent Assessment

1. The process matters as much as the outcome. Employers want to understand which questions candidates ask, how they test hypotheses and how they respond to new information. Traditional take-home assignments are becoming less effective at revealing a candidate's true capabilities when the origin of the answer cannot be established.

2. Assessment is becoming more representative of the actual work environment. If employees will use AI daily to prepare documents, analyse data or write code, a blanket ban on the technology makes practical assessments artificial. Effective assessment should reflect how work is performed today, rather than how it was done in the past.

3. Digital literacy is being divided into technical proficiency and managerial judgement. Knowing how to use ChatGPT is not enough. The real value lies in the ability to delegate tasks appropriately, monitor output quality, account for risks and integrate AI-generated results with professional expertise and business context.

4. Human skills are not disappearing; they are becoming more visible. The faster AI can produce a first draft, the more important critical thinking, ethical judgement, communication, the ability to navigate uncertainty and accountability for decisions become.

Risks of the New Assessment Format

AI-enabled assessment remains an evolving practice. Before implementing it at scale, employers need to address several challenges.

Candidates must be assessed under comparable conditions: they should have the same amount of time, clear instructions and access to an agreed set of tools. Employers also need to establish in advance which data can be shared with external AI models and how the confidentiality of the case will be protected. Assessment criteria should focus on observable candidate behaviours rather than an interviewer's subjective impression of a polished AI interaction.

There is also a methodological risk. Familiarity with a particular AI model may give candidates an advantage that has little to do with their potential job performance. Results from AI-assisted exercises should therefore be considered alongside other sources of evidence, including structured interviews, professional assessments, work samples and the subsequent performance of successful hires.

How to Redesign the Assessment Process

The starting point should not be the choice of AI tool, but an analysis of the role itself. HR teams and hiring managers need to determine:

1. Which tasks employees must be able to perform independently
2. Where the use of AI reflects actual workplace practices
3. Which model errors professionals must be able to recognise
4. Which candidate behaviours can be observed and assessed
5. How candidates will independently defend their final recommendations

A practical assessment model could follow this sequence: independent problem definition → permitted use of AI → validation of the output → introduction of new information or constraints → independent defence of the final decision without AI.

This approach does not replace proctoring, tests or interviews. Instead, it adds another layer of evidence, helping employers understand not only what candidates know but also how they perform in today's technology-enabled workplace.

AI is becoming a participant in the assessment process rather than a substitute for the candidate. The individual therefore remains at the centre of the assessment: their reasoning, judgement, willingness to question assumptions and readiness to take responsibility for the final decision.

Maxim Artemov, Lead Talent Assessment Consultant