McKinsey AI Interview: How Lilli Is Used in Final Rounds and How to Prepare

Based on recent candidate feedback, McKinsey & Company has introduced a live AI-based interview into the U.S. final-round process for selected Business Analyst roles. In addition to traditional case and fit interviews, some candidates are now asked to work directly with Lilli, McKinsey’s internal generative AI platform. In this round, candidates collaborate with Lilli to complete problem-solving and synthesis tasks in real time — closely mirroring how consultants use AI on the job.

This change does not replace classic consulting skills. What has changed is how those skills are evaluated. Candidates are now expected to demonstrate their thinking while working alongside AI — showing ownership of the analysis, the ability to guide the tool effectively, and the judgment to refine and synthesize AI-generated insights.

This article breaks down what the McKinsey AI interview looks like, how Lilli is used in the final round, and — most importantly — how candidates can practice effectively for this new format.

What is Lilli?

Lilli is McKinsey & Company’s proprietary generative AI platform, designed to support consultants as a researcher, synthesizer, and thought partner. It sits at the center of McKinsey’s operational transformation and future way of working.

What can Lilli do?

• Knowledge Aggregation: Lilli draws from over 40 curated internal knowledge sources from McKinsey, including more than 100,000 documents and interview transcripts.
  

• Research and Synthesis: Users can pose questions and receive synthesized insights, source links, and suggestions for relevant internal experts.
  

• Workflow Integration: The platform accelerates research, synthesis, and slide creation, with colleagues reporting up to 30% time savings.
  

• Partner Role: Consultants use Lilli to pressure-test arguments, anticipate client questions, and connect insights across workstreams.

Why does McKinsey do this?

In their report “The Economic Potential of Generative AI,” McKinsey describes AI as a tool that gives professionals “superpowers” by extending human capabilities. Lilli was built to shorten time to insight, increase client impact, and fundamentally rewire how the firm works.

Since its firmwide rollout in July 2023, 72% of colleagues actively use Lilli, generating over 500,000 prompts per month, with consistent reports of significant time savings. As a result, McKinsey expects incoming consultants to be comfortable, thoughtful, and effective in using Lilli.

Expected Candidate Behaviors

Candidates are evaluated on how they work with AI, not just the output. Key behaviors include:

• Drafting clear, structured prompts.
  

• Iterating when initial outputs are imperfect
  

• Applying judgment rather than copying AI responses
  

• Refining outputs before they would be client-ready
  

• Explaining their reasoning, including prompt choices and filtering decisions
  

   

How Lilli is used in the interview

Candidates complete a series of tasks through live interaction with Lilli, which functions as both a research tool and collaborative partner.

Typical tasks include:

• Analytical problem-solving questions
  

• Complex information synthesis
  

• Situational or scenario-based simulations

So how do we practice this?

The change seems to be critical, but do not worry as you only face it in the final round after the classic case interviews and tests. They just want to know if you can use AI in the problem-solving process, and you should know how and when to use Lilli as your partner.

If you can do that, you’re already most of the way there. Below are some steps by step to practice with the AI tool in Case Interview.

Problem Framing (Before AI)

 Start by clearly defining the client objective, decision criteria, and constraints. Decide what kind of output you need and how AI can best support the analysis. This ensures you remain in control of the problem while using AI as a support tool.

Structured Prompting

 Use AI with clear, purpose-driven prompts, such as asking for trends, risks, or a simple structure. Follow up by tightening, organizing, and questioning assumptions. This reflects how consultants interact with Lilli in real engagements.

Hypothesis-Led Interaction

 Approach AI with a working hypothesis and use it to test, refine, or challenge your thinking rather than to generate answers from scratch. This keeps the analysis focused and efficient.

Judgment and Filtering

 Review AI outputs critically by identifying what adds value, what is missing, and what needs refinement before it would be client-ready. Demonstrating judgment matters more than the content itself.

Synthesis and Communication

 Translate AI-assisted analysis into a clear, concise takeaway, acknowledging uncertainty and outlining next steps. Be prepared to explain your logic and decision-making process throughout.

Key baseline

You don’t need many prompts. A strong pattern is:

Frame → One structured prompt → One refinement → One challenge → Human synthesis

If you do that calmly and clearly, you will pass the Lilli round.

Additional Prompt Examples

• Problem Structuring & Scoping

“Create concise talking points to explain this recommendation to a skeptical CEO.”

“Break this problem into 3–4 logical components, without going into detail.”

“What information would a CEO need to make this decision?”

• Hypothesis Generation & Testing

“What are the most likely drivers behind this outcome?”

“What are the most likely drivers behind declining margins in this industry?”
“List 3 hypotheses that could explain this performance gap.”

• Risks, Trade-offs, and Constraints

“What are the main downside risks we should tell to the client?”

“What trade-offs does this option introduce?”

“What could derail execution even if the strategy is sound?”

• Data & Validation

“What data points would you want to validate this conclusion?”

“Where might these assumptions break down?”

“What indicators would confirm or disprove this view?”

• Synthesis & Storylining

“Summarize this into a one-slide executive takeaway.”

“What is the single most important insight here?”

“How would you explain this to a non-technical stakeholder?”

• Stretch / Challenge Prompts

“Argue the opposite recommendation.”

“What are we underestimating?”

“If this fails, what is the most likely reason?”

Looking ahead

McKinsey’s use of Lilli in final-round interviews reflects a broader shift in how candidates are evaluated. The fundamentals remain unchanged — structured thinking, sound judgment, and clear communication — but candidates are now expected to demonstrate these skills while working alongside AI. What matters is not the tool itself, but how effectively you drive the analysis, assess inputs, and synthesize insights.

While McKinsey is currently the most visible firm to formalize a live AI component in its interviews, the wider consulting industry is increasingly embedding internal AI tools into everyday work. As this trend continues, AI fluency is likely to become more relevant in both interviews and on the job across firms, not just at McKinsey.