Management consultancy firm McKinsey & Company announced has announced the launch of Lilli, its new internal generative AI tool aimed at enhancing productivity and creativity for its consultants.
Lilli acts as a virtual research assistant, rapidly synthesizing McKinsey's vast stores of proprietary knowledge to provide tailored insights to client engagement teams. The AI assistant can search over 100,000 internal documents and interview transcripts, as well as external sources, to identify relevant materials and experts on any given topic within seconds.
According to McKinsey senior partner Erik Roth, who led Lilli's development, the tool can answer questions by drawing from the "totality of McKinsey's knowledge." This allows consultants to spend less time on administrative research and more time advising clients. Lilli also surfaces unexpected but apt examples and analogies beyond what consultants may have considered.
The interface includes two main modes - one that sources responses from McKinsey's internal data, and another that taps more generalized AI models. Lilli provides full attribution, linking responses to specific documents and pages.
Built on McKinsey's own technology stack, Lilli leverages large language models like Cohere and tools from OpenAI and Microsoft to enable natural language processing. But McKinsey's layer sits in between to connect users to relevant data sources. The system can also securely analyze client data on McKinsey servers in the future.
The tool was named after Lillian Dombrowski, the first female consultant hired by McKinsey in 1945 who played a key role in shaping the firm's knowledge resources and archives. McKinsey chose the name Lilli to reflect the AI's versatility and thoroughness.
Built using industry-leading AI platforms, Lilli was developed rapidly over the past year by a cross-functional team of over 70 McKinsey experts. It leverages McKinsey's long-standing strengths in knowledge management as well as the advanced AI capabilities of its technology subsidiary QuantumBlack.
Lilli has been in beta testing since June 2023, answering over 50,000 questions already. Consultants report 20% time savings on meeting prep along with higher quality outputs. The tool was named after Lillian Dombrowski, McKinsey's pioneering first female consultant.
Early user feedback on Lilli has been positive. Consultants report time savings of up to 20% on meeting preparation and describe Lilli as a "thought-sparring partner" that improves the quality of their work.
McKinsey plans to continue enhancing Lilli's capabilities and speed. Additional use cases like analyzing client data are also being explored. With other firms racing to deploy AI tools, McKinsey's head start with Lilli provides a key competitive advantage.
By leveraging the institutional knowledge of its thousands of consultants worldwide, combined with leading-edge AI, McKinsey aims to deliver even higher value to its clients. Lilli's rollout across the firm's global workforce later this year will mark a major milestone in McKinsey's AI-powered transformation.