McKinsey has reportedly piloted a hiring assessment in which graduate candidates use its proprietary AI assistant, Lilli, on consulting-style tasks. The exercise is intended to examine how candidates work with AI as well as how they reason about a business problem. It is not evidence that every applicant now takes an AI interview or that AI has replaced McKinsey’s usual recruiting process.
What McKinsey reportedly tested
In a report published in January 2026, the Financial Times described a pilot in which graduate candidates used Lilli while completing consulting-style tasks. CIO’s account characterized the exercise as a way to assess curiosity and judgment, and reported that it was not essential for employment. The reporting describes an additional evaluation, not a simple chatbot quiz or an AI system making hiring decisions.
The task was reportedly designed to resemble consultants’ use of AI: candidates would need to prompt the system, examine and refine its answers, and apply relevant information to a particular client problem. The published accounts do not provide an exact prompt, scoring sheet, or full assessment procedure, so candidates should not assume they can rehearse a known test format.
What Lilli is—and what it does not guarantee
Lilli is McKinsey’s proprietary generative-AI platform for its colleagues. McKinsey says it can search and synthesize material from the firm’s knowledge base and support research and problem-solving. The firm has also described it as an orchestration layer that can coordinate sources and tools, rather than simply a general-purpose chatbot with a McKinsey label. See McKinsey’s introduction to Lilli and its account of lessons from creating the platform.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
McKinsey reports that Lilli was rolled out firmwide in July 2023. Its case study on the platform describes internal use in research, synthesis, and work redesign. Those capabilities do not make any particular answer complete or correct: candidates still need to check whether the response is supported, relevant, and consistent with the facts in the task.
How this fits with McKinsey’s usual recruiting process
McKinsey’s public careers pages continue to describe a role-dependent process that can include application review, assessments, personal-experience interviews, and problem-solving interviews. The firm says its gamified Solve assessment is used for most consulting roles; the exact steps vary by role. Its interview overview and digital-assessment page provide the firm’s public descriptions.
The Lilli exercise is best understood as a reported pilot alongside or within recruiting, not proof that the traditional case interview has disappeared. McKinsey’s public careers material does not establish that all candidates, offices, or job families use Lilli in an assessment, nor does it publish a complete rollout schedule.
What an AI-assisted case could test
The following comparison is an interpretation of the reported pilot, not a published McKinsey scoring rubric. Prompting is only one part of the work: the more consequential question is whether a candidate can use an answer without surrendering judgment to the tool.
Rank #3
- Brand: Burgee Press
- Case in Point 11th Edition: Complete Case Interview Preparation
| A conventional case emphasis | An AI-assisted case may add |
|---|---|
| Structure the business problem | Structure both the problem and the questions asked of the AI |
| Analyze the facts provided | Assess supplied facts alongside AI-generated material |
| Reach a defensible recommendation | Decide what to trust, verify, discard, or adapt for the client |
| Explain the reasoning | Explain the reasoning, relevant uncertainty, and how AI output informed the conclusion |
A strong approach would combine clear problem decomposition with focused prompts, critical review, and concise communication. A polished answer that ignores the client’s circumstances or relies on an unsupported claim can be weaker than a simpler answer whose logic is transparent.
What to do when the AI answer is wrong or generic
Do not treat fluency as proof. If an answer appears questionable, identify the specific claim or assumption that is uncertain, ask for clarification or supporting evidence if the tool allows it, and compare the response with the case facts. Then explain what you would change and why. If the evidence remains incomplete, state the uncertainty rather than presenting the output as established fact.
- Check whether the answer addresses the question actually asked.
- Verify calculations and claims that drive the recommendation.
- Look for missing facts, unstated assumptions, and alternative explanations.
- Separate evidence from inference, and adapt general advice to the client’s market or operating context.
- Make and defend your own recommendation; do not ask the tool to make the decision for you.
McKinsey’s descriptions of Lilli explain its intended knowledge and synthesis capabilities, not an assurance that every response is reliable in every context. The public reporting does not establish a formal rubric for how the pilot handles AI errors.
When candidates may use AI
McKinsey draws a clear line between preparation and assessment performance. Its candidate guidance says AI may be used for preparation such as polishing a résumé, practicing interview questions, or explaining concepts and frameworks. It says candidates may not use AI to invent or exaggerate achievements, generate real-time answers in an interview, or use it in an assessment unless specifically permitted.
Best Value
- Job Work Order Book for Small Business: the package consists of 2 work order forms book, which 33 sets per book, give you 66 total sets three part carbonless invoices, which sufficient quantity is ideal for your businesses.
- 3 Part Carbonless Work Order Forms: the job work order book measures about 5-9/16 x 8-7/16 Inches, and the 3 part carbonless could give customers one copy, retain the additional two copies for your records.
- Wraparound Back Cover: the job work order book wraparound back cover prevents write-through between sets, when you fill in the orders, you can open the folded back cover and place it under the set of orders you want to fill in, so as to prevent the written content from rubbing onto the subsequent copy of the purpose work order forms book.
- Custom Invoice Receipt Book: each general purpose job work invoice receipt has a blank space up top to give you room for a company stamp.
- Job Work Order Forms Book Usages: our job work order book is an essential supply for small business owner or entrepreneur, each work order stay organized.
The firm’s assessment-integrity expectations also tell candidates not to use applications, websites, generative AI, calculators, or prepared notes during assessments or interviews unless expressly allowed. The firm asks candidates to disable previously installed AI software that could record or assist during an interview.
So a reported Lilli exercise would be an explicitly authorized exception for that exercise—not permission to use a personal ChatGPT account, browser extension, transcription tool, or other assistant in a different interview. Follow the invitation and assessment instructions for the specific role; if permission is unclear, ask the recruiter before starting.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to prepare without relying on a leaked test format
- Build the fundamentals first. Practice structuring ambiguous problems, forming hypotheses, interpreting quantitative information, synthesizing findings, and communicating recommendations. Prepare personal-experience examples as well; an AI exercise would not replace every part of the process.
- Practice reviewing AI output. For a short case you have already attempted, ask an AI tool for an alternative structure or overlooked considerations. Identify unsupported assumptions, errors, or generic advice before using any of its suggestions.
- Use follow-up questions to investigate, not to decorate. Ask what assumptions are being made, which facts could change the recommendation, what evidence is missing, or what a skeptical client might challenge.
- Translate the result into a client-specific answer. Explain which points matter for the client’s geography, customer segment, or operating model, and which do not. Keep the recommendation tied to the case facts.
- Rehearse explaining your judgment. Be ready to say what you accepted, rejected, or changed in an AI response and why. This is a preparation exercise, not a claimed replica of McKinsey’s pilot.
- Use only tools the real assessment authorizes. If an assessment provides an employer-approved AI tool, work within that environment and its rules. Do not bring in a personal assistant or prepared notes unless the instructions expressly permit them.
McKinsey also offers a separate AI tool for interview preparation
McKinsey’s Prep Partner is a firm-branded preparation resource, distinct from the reported Lilli hiring pilot. Its page warns that AI-generated material may contain inaccuracies or omissions. Its availability and terms should be checked on the tool’s own page; its existence does not authorize use during a live interview or assessment.
What remains unconfirmed
Public reporting and McKinsey’s public recruiting pages do not establish which offices or degree programs took part, how candidates accessed Lilli, the precise scoring method, whether human reviewers examined the interaction, or whether the pilot expanded. They also do not establish whether the exercise is mandatory for any particular role or location. Applicants should rely on the instructions attached to their own application rather than infer a universal requirement from the pilot.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesWhy the experiment matters beyond prompting
The pilot points to a possible shift in what employers want to observe: not only whether someone can produce an answer unaided, but whether that person can use AI while retaining responsibility for the result. That can make a task more representative of AI-enabled work, but it also raises practical challenges around consistent tool behavior, candidate familiarity, accessibility, and fair comparison. The reported experiment is not, by itself, evidence that McKinsey or other consulting firms have adopted a common industry-wide hiring standard.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
