McKinsey Lilli: a generative AI platform for firm-wide knowledge

McKinsey rolled out its generative AI platform Lilli across the firm in July 2023. According to McKinsey, 72% of its people are active on the platform, it answers more than 500,000 prompts a month, and employees report saving up to 30% of the time they spend searching for and synthesizing knowledge.

At a glance

Item	Detail
Company	McKinsey & Company
Industry	Consulting
Use cases	Business intelligence, process optimization
Functions	Business intelligence, process optimization
Reported results	Up to 30% less time gathering and synthesizing information; 20% better content quality and accuracy in blind review; 500,000+ prompts a month. No dollar savings, hours or team-size changes disclosed
Moments rating	S-tier quick win
AI use-case score	95

Case details

Background

Knowledge and expertise are McKinsey’s core assets. Lilli connects more than 40 of the firm’s knowledge sources and capabilities, and is named after Lillian Dombrowski, the first woman McKinsey hired as a professional, in 1945. Early prototype testers were already using it for data analysis, planning and creative problem solving, so McKinsey chose to build Lilli as a platform for generative AI applications rather than a search tool.

What they did

• Ran it as a transformation program. McKinsey applied its own “Rewired” framework with explicit goals. Senior partner Rodney Zemmel led the firm’s overall AI transformation; senior partner Erik Roth led Lilli’s development.
• Changed the organization. A new operating model brought together teams that had not worked closely before.
• Put security first. Risk and security were priorities from day one. Fortune reports that staff may enter confidential client data into Lilli, but not into tools such as ChatGPT.
• Kept adding capabilities. A year in, the focus shifted to agents that automate specific time-consuming tasks and to slide creation. Fortune also mentions a “Tone of Voice” tool for checking reports, and growing use of Lilli to draft proposals.
• Drove adoption. Ongoing training, a community of generative AI enthusiasts and skill building raised usage.

Results and data

• 72% of employees active on the platform (McKinsey case page); Fortune reported in June 2025 that more than 75% use it monthly.
• More than 500,000 prompts answered each month.
• Up to 30% time saved on searching and synthesizing knowledge.
• A 20% improvement in content quality and accuracy, based on a five-point blind review of work produced with and without Lilli.

McKinsey acknowledges there is still work to do on daily usage. Neither source discloses hours or money saved, or any change in team size.

What to take from it

• Consolidate knowledge sources first. The value comes from putting scattered internal material behind one entry point, so audit and clean your sources before building a knowledge base assistant.
• Treat it as organizational change. Clear goals, a senior owner and a cross-team operating model, not a project handed to IT.
• Measure quality with blind review. Time saved is only half the story; a controlled blind comparison of output quality is a method worth copying.

Limits

• The 72% and 30% figures are self-reported by McKinsey, and “up to 30%” is the top of employee feedback, not an average.
• McKinsey has deep, structured research and project archives. Companies with thin documented knowledge are unlikely to see the same results.
• Industries handling confidential client data need to solve permissions and security first, as McKinsey did, before scaling up.

Sources

Original source: McKinsey · Rewiring the way McKinsey works with Lilli Additional source: Fortune · McKinsey’s AI use for slides and proposals

This is an industry research entry: a publicly documented third-party AI case, not a Moments client case. For projects Moments delivered itself, see client cases.