For decades, McKinsey's expertise lived in siloed practice groups, disconnected data sources, and the heads of senior partners, brilliant, but unreachable at the moment a consultant needed it most. Kitti Lakner, the creator of Lilli, sat down with Box's Jon Herstein to explain how large language models finally made it possible to bring all of that knowledge together in a federated, consumable way. Lilli is not a search engine. It is an AI-driven sounding board, a thought partner, and what Kitti calls an "Iron Man builder" for McKinsey's consultants. Before Lilli, finding the right expert for an engagement took days of manual outreach. Now a consultant can ask Lilli who the best expert is on a given topic, get a ranked answer grounded in McKinsey's full knowledge graph, and walk into a client meeting better prepared, in a fraction of the time. The conversation covers how Lilli handles permissioned access across practice groups (M&A content requires certification; periscope assets require engagement alignment), how AI levels the playing field for junior consultants while multiplying the output of senior ones, and why one associate partner in aerospace is now doing the work of a team of three or four using today's AI capabilities. Kitti also addresses the harder questions: How do you prevent problem-solving skills from atrophying when AI handles the heavy lifting? How do you drive adoption in a 100-year-old organization? Her answer: change management costs more than the technology. For every dollar spent on tech, plan to invest four in change management. The firm rebuilt its entire upskilling curriculum, from onboarding through senior partner orientation, and found that adoption accelerates fastest when senior leaders model the behavior themselves. One senior partner learned a vibe coding tool on a Friday and had his entire team using it by Wednesday. Kitti shares how McKinsey started building Lilli in late 2022 when no enterprise SaaS option existed, always with the intention of swapping in best-in-class components as the market matured. She also addresses AI risk: with citizen development expanding rapidly, ensuring colleagues understand how to separate code from data, and how to move fast without creating security exposure is a daily priority. Chapters: 00:00 — Opening: The problem-solving atrophy question 00:21 — What is Lilli and why did McKinsey build it? 01:26 — How knowledge was siloed before Lilli 03:17 — Access vs. awareness: the real bottleneck 04:16 — Permissioned knowledge and practice-specific access 05:45 — Consultant workflow before and after Lilli 08:29 — Does AI level the playing field or widen the gap? 10:05 — How AI multiplies senior experts (the 10x effect) 11:43 — Adoption surprises and the pace of change 13:43 — Upskilling: are organizations underestimating the work? 17:10 — The junior-to-senior knowledge pipeline question 18:08 — Preventing problem-solving atrophy 20:47 — Change management costs more than the technology 26:59 — Build vs. buy: how McKinsey's approach evolved 29:45 — Knowledge and data as organizational moats 31:10 — What most organizations get wrong about AI adoption 32:28 — Measuring value when the benefit isn't just time saved 35:01 — When is an AI system ready to move from POC to production? 37:19 — Closing FAQs: Q: What is Lilli and what does it do for McKinsey consultants? A: Lilli is McKinsey's internal AI platform that provides consultants with conversational access to the firm's collective knowledge, curated content, and proprietary assets to accelerate research and client preparation. Q: Does giving junior consultants AI access reduce the value of senior expertise? A: No. AI levels the playing field for foundational tasks but multiplies senior experts rather than replacing them, amplifying their output and expertise up to tenfold. Q: How did McKinsey drive adoption of Lilli across the firm? A: McKinsey shifted focus from prompt training to use cases and incentives, rebuilt its upskilling curriculum, ran team-embedded learning, and leveraged visible tool usage by senior leaders as the primary adoption driver. Q: How does McKinsey handle the risk that AI will cause problem-solving skills to atrophy? A: McKinsey redesigned workflows so AI automates mechanical tasks (like PowerPoint formatting) while critical problem-solving discussions and arguments remain entirely human-led. Q: What is McKinsey's view on build vs. buy for AI tools? A: McKinsey built Lilli in 2022 when no enterprise SaaS option existed, but designed it to swap in best-in-class third-party components (like OpenAI's Assistants API and external rerankers) as the market matured. Q: What does Kitti Lakner mean when she says change management costs more than the technology? A: For every dollar spent on technology, organizations should invest four in change management. Driving behavioral change, aligning incentives, and getting leaders to model new tools is harder and slower than deployment.

 43181      31