The McKinsey Lilli interview is becoming an important topic for candidates preparing for the McKinsey final round interview, especially as AI becomes part of consulting work and recruiting. Lilli is McKinsey’s proprietary AI platform, and the interview pilot shows how candidates may be asked to use AI during a case-style problem. The focus is not just on the answer, but on prompt quality, judgment, output validation, and synthesis. In this article, we will explore what Lilli is, how the interview works, what McKinsey may evaluate, and how to prepare using publicly available AI tools.
TL;DR – What You Need to Know
The McKinsey Lilli interview tests how candidates use AI to structure, validate, and synthesize case insights in final rounds.
- Lilli is McKinsey’s internal generative AI platform for knowledge search, synthesis, and expert discovery.
- The interview pilot asks candidates to collaborate with Lilli on a case style business problem.
- McKinsey evaluates prompt quality, structured thinking, judgment, output validation, and client ready synthesis.
- Candidates should prepare with public AI tools while maintaining strong case interview fundamentals.
- Lilli reflects a broader MBB shift toward AI enabled consulting work.
What Is the McKinsey Lilli Interview?
The McKinsey Lilli interview is an AI enabled final round pilot where candidates use Lilli, McKinsey’s internal generative AI tool, to work through a case style business problem. The exercise tests how candidates structure prompts, assess AI outputs, and turn findings into a clear client ready recommendation.
In practical terms, the McKinsey Lilli interview is not a replacement for the traditional case interview or McKinsey PEI. It is an additional pilot format reported in select final round processes, where candidates collaborate with Lilli during a consulting style problem rather than solving everything independently.
The core idea is simple: McKinsey wants to observe how candidates think when AI becomes part of the problem solving workflow. You are still responsible for structuring the issue, identifying the right questions, interpreting the output, and explaining what the client should do next.
A strong candidate does not treat Lilli as the answer. A strong candidate treats Lilli as an input.
In the interview, you may be expected to show:
- Clear problem structuring before using the tool
- Focused prompts that guide the analysis
- Judgment when reviewing AI generated responses
- Ability to challenge incomplete or generic outputs
- Synthesis into a practical recommendation
- Communication that sounds client ready, not tool dependent
This matters because Lilli is already part of how McKinsey works internally. McKinsey describes Lilli as a generative AI platform that helps colleagues search and synthesize the firm’s knowledge, including more than 100,000 documents and interview transcripts across curated knowledge sources.
For candidates, the key takeaway is that the McKinsey Lilli interview is less about technical AI expertise and more about consulting judgment. McKinsey is not simply checking whether you can write prompts. It is checking whether you can use AI while staying in control of the analysis.
That means your preparation should still start with case fundamentals. You need to break down ambiguous problems, form hypotheses, interpret data, and communicate recommendations clearly. Lilli may support the work, but you remain accountable for the thinking.
What Is the McKinsey Lilli Platform?
The McKinsey Lilli platform is McKinsey’s internal generative AI system for searching, synthesizing, and applying firm knowledge. It helps consultants access curated documents, identify relevant insights, and find experts more quickly, which makes it central to understanding why AI enabled problem solving is entering consulting interviews.
McKinsey introduced Lilli as a tool for colleagues to search and synthesize the firm’s knowledge base. The firm says Lilli brings together more than 40 curated knowledge sources, more than 100,000 documents and interview transcripts, and experts across 70 countries.
Lilli is named after Lillian Dombrowski, the first professional woman hired by McKinsey in 1945. McKinsey says Dombrowski later became controller and corporate secretary, helped create the firm’s archives, and contributed to research that supported new practice areas.
For candidates, this background matters because Lilli is not a public chatbot. It is a proprietary AI tool designed around McKinsey’s own knowledge and consulting workflows.
The McKinsey Lilli platform can support consultants by helping them:
- Search firm knowledge more efficiently
- Summarize relevant documents and transcripts
- Identify five to seven relevant pieces of content
- Surface experts connected to a topic
- Support early project planning and research
- Pressure test arguments before client discussions
McKinsey has described Lilli as a search and synthesis platform, not a replacement for consultant judgment. The user still needs to ask the right question, evaluate the output, and apply the insight to the client context.
That distinction is important for the interview. The value of Lilli is not that it gives a perfect answer. The value is that it can accelerate the search for relevant inputs while the consultant remains responsible for analysis, judgment, and recommendation quality.
Kickstart Your Consulting Prep Journey?
Click the image below to get your free Consulting Starter Pack
.png)
How the McKinsey Lilli Interview Works
The McKinsey Lilli interview works as an AI enabled case exercise where final round candidates use Lilli during a business problem. The candidate is expected to frame the case, prompt the tool, evaluate the response, refine the analysis, and present a clear recommendation.
The reported format places Lilli inside the case process rather than outside it. You are not simply asked whether you know what AI is. You are asked to show how you use AI while solving a consulting problem.
A typical flow may look like this:
- Receive a business scenario from the interviewer
- Clarify the client objective and constraints
- Build an initial issue tree or hypothesis
- Use Lilli to explore relevant drivers or examples
- Review the AI generated output critically
- Ask follow up prompts to refine the analysis
- Synthesize the findings into a recommendation
- Explain what you would do next for the client
The most important point is control. The candidate should lead the problem solving process, not let the tool dictate the path.
For example, a weak approach would be to ask Lilli for a full answer immediately and repeat that output. A stronger approach would be to define the business problem first, ask Lilli targeted questions, compare the response against your structure, and state what you accept or reject.
The interview is therefore interactive. You are cycling between human judgment and AI support.
That makes communication especially important. The interviewer needs to hear why you are asking a prompt, what you think of the response, and how it changes your recommendation.
A good candidate might say:
- “I will first isolate the growth drivers before asking Lilli for relevant market patterns.”
- “This output is helpful, but it is too broad for the client’s region.”
- “I would not rely on this conclusion yet because it lacks customer segment detail.”
- “The most useful insight is the margin pressure point, so I will build the recommendation around that.”
This is still a consulting case interview at its core. The AI changes the workflow, but the candidate’s responsibility stays the same: structure the ambiguity, use evidence carefully, and give the client a practical answer.
What McKinsey Evaluates in AI Enabled Cases
McKinsey evaluates AI enabled cases by looking at how candidates combine structured thinking, prompt quality, judgment, and synthesis. The strongest candidates use AI as a thinking partner, challenge its outputs, and translate useful insights into a client ready recommendation.
The McKinsey AI interview is not mainly a test of technical AI knowledge. You do not need to know how a language model is trained or how McKinsey built Lilli. You need to show that you can use AI responsibly inside a business problem.
The most important evaluation areas are likely to include:
- Problem structuring: Can you break the case into logical drivers before prompting?
- Prompt quality: Can you ask focused questions that produce useful inputs?
- udgment: Can you tell when an answer is generic, incomplete, or unsupported?
- Iteration: Can you refine your approach as new information appears?
- Synthesis: Can you turn scattered outputs into a clear recommendation?
- Communication: Can you explain your reasoning in a concise, client ready way?
The biggest difference from a standard case is that your process becomes more visible. In a traditional case, the interviewer sees your structure, math, and recommendation. In an AI enabled case, the interviewer also sees how you guide a tool and respond to uncertainty.
That creates both opportunity and risk.
It is an opportunity because strong candidates can show maturity beyond basic case mechanics. You can demonstrate that you know when to use AI, when to ignore it, and when to ask for better evidence.
It is a risk because weak candidates may hide behind the tool. If your prompts are vague, your critique is shallow, or your final recommendation simply repeats the AI output, the interviewer may question your consulting judgment.
Think of Lilli as an analyst on your team. A consultant does not accept every analyst output without review. A consultant checks relevance, pressure tests assumptions, and decides what belongs in the final client message.
That is the mindset this exercise rewards.
How Lilli Is Reshaping Consulting Work
McKinsey Lilli is reshaping consulting work by making knowledge search, synthesis, and early problem solving faster. Consultants can use the tool to access firm knowledge, identify relevant sources, and prepare for client work, while still relying on human judgment for context and recommendations.
McKinsey says Lilli was built to help colleagues access the firm’s knowledge more efficiently and spend more time on problem solving, coaching, capability building, and client impact. The platform can scan knowledge sources, summarize key points, include links, and identify relevant experts.
That changes how junior consultants may work. Tasks that once required long searches across documents, expert lists, and prior project materials can become more direct.
Instead of asking, “Where do I start?” a consultant can ask better questions earlier:
- What have we seen in similar client situations?
- Which cost drivers usually matter in this industry?
- What examples could pressure test this hypothesis?
- Which internal experts understand this topic?
- What risks might weaken this recommendation?
The output still needs review. McKinsey itself notes that Lilli is part of a broader workflow involving prompt engineering, content validation, and secure scaling.
For candidates, the practical implication is clear. If consulting work increasingly involves AI supported research and synthesis, then recruiting can reasonably start testing whether you can work in that environment.
This does not mean consultants stop thinking. It means the baseline expectation changes.
You may need to become faster at:
- Defining the question before searching for answers
- Asking targeted prompts
- Testing AI generated claims against business logic
- Separating useful insight from generic language
- Turning rough output into executive ready communication
That is why the McKinsey Lilli interview matters. It reflects a real shift in consulting workflows, not just a new interview novelty.
How to Prepare for the McKinsey Lilli Interview
To prepare for the McKinsey Lilli interview, practice case interviews with publicly available AI tools while staying in control of the analysis. Focus on structuring the problem first, writing specific prompts, challenging outputs, and converting insights into a concise recommendation.
Candidates are not expected to have access to Lilli before the interview. Preparation should therefore focus on transferable skills, not tool specific shortcuts.
A practical preparation workflow looks like this:
1. Start with normal case fundamentals
Before using AI, make sure you can solve a case without it. You still need market sizing, profitability logic, issue trees, hypothesis driven thinking, and synthesis.
If your case basics are weak, AI will not fix the problem. It may make the weakness more obvious because your prompts will lack direction.
2. Practice AI assisted case drills
Use a public AI tool to simulate the workflow. Give yourself a business problem, build a structure, then ask the tool for targeted inputs.
For example:
- “List possible revenue drivers for a premium fitness chain entering a new city.”
- “Give three risks that could weaken this market entry recommendation.”
- “Challenge this profitability hypothesis from the client’s perspective.”
- “Identify missing information before making a recommendation.”
The goal is not to collect perfect prompts. The goal is to practice managing AI as part of your thinking process.
3. Evaluate every output
After each response, ask yourself:
- Is this specific to the case?
- Does it answer the actual question?
- What assumption is it making?
- What evidence is missing?
- What would I ask next?
- What would I ignore?
This is where judgment shows. Many AI responses sound polished, but polished does not always mean useful.
4. Synthesize out loud
After each AI interaction, summarize what changed.
For example:
- “This supports the pricing hypothesis, but I still need margin data.”
- “The output suggests customer retention may matter more than acquisition.”
- “I would deprioritize this point because it is not specific to the client’s market.”
This habit helps you avoid getting stuck in endless prompting. In the interview, the final answer still matters.
5. Practice client ready recommendations
End every drill with a short recommendation. Use a simple structure:
- Recommendation
- Two or three reasons
- Key risk
- Next step
A strong final answer might sound like this:
“I recommend entering the market through a limited pilot in two metro areas because demand appears concentrated, acquisition costs can be tested quickly, and the client can limit upfront investment. The main risk is weak retention after trial offers, so the next step is to validate repeat purchase behavior before scaling.”
That is the kind of synthesis AI can support, but you need to deliver it.
Common Mistakes Candidates Should Avoid
Candidates should avoid treating AI as the decision maker, accepting outputs without challenge, and failing to synthesize. In an AI enabled case, the interviewer is looking for judgment, not dependency, so your role is to guide the tool and own the recommendation.
The first mistake is asking the tool to solve the entire case. This weakens your position because it shows you are outsourcing the thinking.
A better approach is to structure the problem first, then use AI for targeted support.
Weak prompt:
- “Solve this case.”
Stronger prompt:
- “Identify three cost drivers that could explain declining margins in a regional grocery chain, and state what data would confirm each one.”
The second mistake is accepting AI output too quickly. AI can be confident and incomplete at the same time.
You should always look for:
- Missing assumptions
- Unsupported claims
- Generic statements
- Overly broad recommendations
- Lack of client context
- Contradictions with the case facts
The third mistake is over prompting. Some candidates may keep asking follow up questions without moving toward a conclusion.
That can signal poor prioritization. Consultants need to know when enough information exists to make a reasonable recommendation.
The fourth mistake is using AI language without translating it. If the output sounds abstract, you need to convert it into business terms.
For example, “enhance customer experience” is too vague. A more useful client version would be, “reduce onboarding friction by shortening sign up from five steps to three and testing whether conversion improves.”
The fifth mistake is forgetting the interviewer. You should narrate your thinking, not silently interact with the tool.
A strong candidate explains:
- Why they are asking a prompt
- What they learned from the output
- What they trust or do not trust
- How the insight affects the recommendation
The safest mindset is simple: AI can help you think faster, but it should not think for you.
Why Lilli Signals an MBB Wide AI Shift
Lilli signals an MBB wide AI shift because top consulting firms are building internal AI tools for knowledge retrieval, synthesis, slide creation, and case support. The recruiting implication is that candidates may increasingly need to show AI fluency alongside traditional case interview skills.
McKinsey is not alone in using AI inside consulting workflows. Bain says it built Sage, an in house AI platform powered by ChatGPT, to help consultants find case insights, summarize research, and connect with internal experts.
BCG has also developed internal generative AI tools. Business Insider reported that one BCG tool, Deckster, supports slide creation and review, while other tools support brainstorming and knowledge workflows.
The important point is not that every firm will use the same interview format. They probably will not.
The important point is that consulting work is becoming more AI enabled across research, synthesis, communication, and delivery. As that happens, recruiting may place more weight on how candidates use AI responsibly.
For McKinsey candidates, this means the Lilli pilot should be treated as a signal. It shows that the firm may care about:
- How you ask questions
- How you handle imperfect information
- How you validate AI generated output
- How you communicate recommendations
- How you combine speed with judgment
At the same time, candidates should avoid overreacting. Public information still indicates that traditional case interviews and fit interviews remain central to consulting recruiting. The Lilli exercise should not replace your case preparation, PEI preparation, or mental math practice.
The better interpretation is balance. Prepare for AI enabled consulting work, but do not abandon the fundamentals that consulting interviews have always tested.
Frequently Asked Questions
Q: What is the McKinsey Lilli interview?
A: The McKinsey Lilli interview is an AI-enabled final round pilot where candidates use Lilli to work through a case-style business problem. It tests structured thinking, prompt quality, output validation, and synthesis into a client-ready recommendation.
Q: How does the McKinsey Lilli interview work?
A: The McKinsey Lilli interview works by asking candidates to frame a consulting case, collaborate with the McKinsey Lilli platform, review AI outputs, and refine their recommendation. The interviewer looks at how candidates guide the tool, not just the final answer.
Q: Is the McKinsey Lilli interview evaluative?
A: The McKinsey Lilli interview has been reported as a final round interview pilot rather than a universal replacement for McKinsey’s standard case and PEI process. Candidates should still prepare seriously because it signals growing importance around AI fluency.
Q: How do you prepare for the McKinsey Lilli interview?
A: To prepare for the McKinsey Lilli interview, practice case interviews with public AI tools while staying in control of the analysis. Focus on structuring, writing targeted prompts, challenging AI outputs, and delivering a clear recommendation.
Q: What skills does McKinsey test in the Lilli interview?
A: McKinsey tests structured thinking, prompt quality, business judgment, AI collaboration, and synthesis in the Lilli interview. Strong candidates challenge AI outputs and translate useful insights into a practical, client-ready recommendation.
