Lilli by McKinsey: Your AI Consultant for the Future
McKinsey and Company, the global consulting titan, has taken a quantum leap into the realm of artificial intelligence, introducing its groundbreaking generative AI tool, Lilli. With nearly half of its 30,000-strong workforce already leveraging AI, McKinsey’s proprietary creation promises to redefine how consulting services are delivered.
Named after Lillian Dombrowski, a trailblazing figure who joined McKinsey in 1945 as the firm’s first woman in a professional role, Lilli isn’t just another AI experiment. This sophisticated chat application has emerged from McKinsey’s “ClienTech” team, led by Chief Technology Officer Jacky Wright. Lilli transforms consulting workflows by analyzing over 100,000 documents and interviews, providing custom insights, recommendations, data, plans, and expert connections.
Erik Roth, a senior partner at McKinsey, elaborates on the concept: “If you could ask the totality of McKinsey’s knowledge a question, and [an AI] could answer back, what would that do for the company? That’s exactly what Lilli is.”
Lilli’s beta phase, initiated in June 2023, has seen resounding success. In addition, around 7,000 employees have already experienced the tool’s prowess as a “minimum viable product” (MVP), witnessing drastic reductions in research and planning time. Furthermore, Roth reports Lilli answered 50,000 questions in 2 weeks, with 66% of users returning for more weekly.
So, how does Lilli work?
The user-friendly interface, reminiscent of popular AI tools like ChatGPT, features a text entry box for user queries and generates responses in a chronological chat format. However, Lilli’s uniqueness shines through its side panel filled with saved prompts that users can personalize. Roth reveals that categorized prompts are on the horizon, enhancing usability.
Two distinct tabs further augment Lilli’s capabilities. “GenAI Chat” taps into a vast language model, while “Client Capabilities” draws insights from McKinsey’s extensive database. Notably, Lilli goes the extra mile by providing transparent source attribution, a feature that excites clients. It links responses to specific pages, underpinning its credibility.
Lilli is suitable for performing a wide range of tasks. From preliminary research on client sectors to drafting project implementation strategies, McKinsey envisions Lilli as an indispensable consultant’s aide. During a demonstration, Lilli efficiently identified internal experts, forecasted clean energy trends, and even planned a ten-week energy plant construction.
Despite some responses being marginally slower than commercial counterparts, McKinsey prioritizes quality over speed. Roth reassures that performance updates are ongoing. Importantly, McKinsey is considering adding a secure client data analysis feature to Lilli, ensuring compliance and usefulness.
Under the hood, it leverages existing language models, including those by McKinsey partner Cohere and OpenAI, acting as an intermediary between users and data. Roth’s vision extends beyond McKinsey’s internal use. This suggests that McKinsey could white-label or externally offer it to other organizations.
McKinsey’s Lilli is set to redefine consulting, harnessing AI to amplify human expertise in our rapidly evolving tech landscape. Moreover, as Roth aptly states, “I personally believe that every organization needs a version of Lilli.”
