McKinsey's Secret AI Tool Lilli — What It Does

McKinsey Built an AI Tool Nobody Was Supposed to See

McKinsey & Company — the most secretive consulting firm on the planet — has been quietly building an internal AI tool called **Lilli** since late 2023. Until now, details were sparse. But a combination of leaked internal documents, employee LinkedIn posts, and conference presentations has given us a surprisingly complete picture of what Lilli actually does.

And it's a big deal — not just for consulting, but for anyone who does knowledge work.

---

What Is Lilli?

Lilli is McKinsey's proprietary AI knowledge management and research synthesis platform. Named after Lilli Boulanger (a nod to McKinsey's internal culture of classical references), the tool is designed to do one thing exceptionally well: **turn McKinsey's massive internal knowledge base into instant, actionable insights.**

Think of it as a private, enterprise-grade version of Perplexity — but trained on decades of McKinsey's proprietary research, frameworks, client deliverables, and expert interviews.

Key stats:

---

What Lilli Actually Does

1. Research Synthesis at Scale

The core use case: a consultant asks Lilli a complex question like *"What are the key success factors for digital transformation in European manufacturing?"* and gets a synthesized answer pulling from dozens of internal reports, with citations.

What makes it different from ChatGPT:

2. Expert Network Matching

Lilli doesn't just find documents — it finds **people**. When a consultant needs domain expertise, Lilli identifies the right internal expert based on their past project work, published research, and skill profiles.

This solves one of consulting's biggest problems: the "who knows what" challenge in a 45,000-person organization.

3. Client Deliverable Generation

Here's where it gets interesting. Lilli can draft sections of client deliverables — slide narratives, executive summaries, market sizing frameworks — using McKinsey's house style and methodology.

Reported capabilities:

4. Knowledge Gap Identification

Lilli flags when McKinsey doesn't have strong internal data on a topic. This helps teams identify where they need primary research, expert interviews, or new analysis — before the client asks.

---

The Technology Behind Lilli

Based on available information, Lilli runs on:

The system is reportedly hosted on Microsoft Azure with strict data residency controls for different regions.

---

Why This Matters Beyond McKinsey

The Consulting Productivity Gap

McKinsey's investment in Lilli signals something important: **even the best-resourced knowledge workers need AI to stay competitive.** If McKinsey — with 45,000 consultants and decades of institutional knowledge — needs AI to synthesize its own research, what about everyone else?

What You Can Replicate Today

You don't need McKinsey's budget to build something similar. Here's how to approximate Lilli's core capabilities with available tools:

Research Synthesis:

Knowledge Management:

Deliverable Generation:

Expert Finding:

---

The Competitive Implications

Every major consulting firm is building its own version:

The firms that integrate AI deepest into their workflows will deliver faster, cheaper, and more data-driven insights. The ones that don't will lose talent and clients.

---

Keep Reading

JPMorgan's AI Now Writes Research Reports (2026)6 AI Tools That Cut Consulting Deliverables 60%

What This Means for Your Work

Whether you're a consultant, analyst, researcher, or knowledge worker, McKinsey's Lilli represents the direction all professional work is heading:

---

Bottom Line

McKinsey's Lilli isn't just a chatbot — it's a knowledge infrastructure that turns decades of consulting expertise into on-demand intelligence. While you can't use Lilli itself, you can build similar capabilities using available AI tools.

The firms and professionals who invest in AI-powered knowledge management now will have a compounding advantage. The ones who treat AI as a novelty will get left behind.

Start building your own "Lilli" today — even if it's just Notion AI plus Perplexity Pro plus Tome. The stack matters less than the habit of using AI to synthesize, not just search.