Case Study: McKinsey’s GenAI Platform

Developed initially as a knowledge extraction and synthesis tool, Lilli now coordinates various forms of knowledge across McKinsey and beyond, helping consultants deliver greater impact for clients. Lilli’s development showcases how McKinsey approached the AI transformation process with an emphasis on user-centric design, iterative testing, and controlled adoption.

Key Takeaways

• Lilli combines multiple models and technologies to coordinate McKinsey’s knowledge base securely and efficiently.
• Built through ethnographic research, Lilli focuses on solving specific user problems in four domains: high-performing team enablement, client development with AI-augmented strategies, delivery of distinctive client service, and post-engagement quality communications
• McKinsey promoted Lilli’s adoption through leadership role modeling, comprehensive training, and integration into daily workflows.
• Lilli allows consultants to focus on activating insights rather than spending time on analytics, enhancing efficiency.
• Lilli signals a shift toward AI-driven consulting, with greater diversity, tech enablement, and human impact.

Approach

The development of Lilli began as a small-scale experiment with a team of just four people. McKinsey adopted an iterative and user-centric approach, conducting ethnographic research to identify and address specific consultant needs. The focus remained on four key domains: enabling high-performing teams, developing clients with AI-augmented strategies, delivering distinctive client service, and maintaining high-quality communications post-engagement.

McKinsey ensured that every feature developed was tied directly to solving a user problem, keeping the user experience central to the process. Integration with McKinsey’s culture was also prioritized, as leaders encouraged Lilli’s adoption through role modeling and consistent usage in team meetings, asking, “Have you asked Lilli?” These efforts were complemented by structured training programs, user groups, and iterative feedback loops to refine the platform continuously.

Implementation

Lilli’s development relied on a sophisticated technology stack that combined large and small models to deliver precise and reliable outputs. Unlike tools based solely on retrieval-augmented generation (RAG), Lilli acts as an orchestration layer, coordinating diverse knowledge sources while maintaining security and efficiency. The rollout followed a controlled approach, starting with a group of 2,500 McKinsey colleagues to gather feedback and refine the platform.

Gradual scaling allowed for iterative testing and the creation of Lilli evangelists, who promoted adoption across the firm. Training sessions, user communities in ten offices, and built-in guides further facilitated adoption. Key use cases emerged, such as knowledge synthesis, AI-augmented consulting, and the McKinsey Tone of Voice Agent, which ensures high-quality, professional writing tailored to McKinsey standards. These capabilities enable consultants to spend less time on repetitive tasks and more time delivering actionable insights to clients.

Results

Lilli’s widespread adoption has made it one of the most-used tools at McKinsey, reflecting its significant value to consultants. The platform has improved productivity by reducing the time spent on knowledge extraction and synthesis, allowing consultants to focus on activating insights for clients. Tools like the McKinsey Tone of Voice Agent enhance the quality of communications, particularly for non-native English speakers, ensuring clarity and professionalism.

Beyond productivity, Lilli has contributed to a cultural shift within McKinsey, where consultants are increasingly tech-enabled and better positioned to deliver impact. This transformation has also enabled more diverse and empathetic consulting practices by upskilling professionals and broadening the talent pool. Overall, Lilli has elevated the quality and efficiency of McKinsey’s services, solidifying its role as a key enabler of AI-powered consulting.

Challenges and Barriers

Despite its success, Lilli’s development faced several challenges. Data readiness was a major barrier, as effective AI tools require well-structured, tagged, and curated data to function optimally. Adoption resistance also emerged in the early stages, requiring sustained leadership efforts and cultural changes to encourage usage.

Another challenge was balancing the pace of development with user learning. Building new capabilities often outpaced the time needed to gather meaningful user feedback, leading McKinsey to adopt alpha and beta testing strategies to prioritize learning. Additionally, maintaining AI safety and addressing biases required robust systems to ensure ethical usage and data privacy.

Future Outlook

The future of consulting at McKinsey will be increasingly shaped by AI tools like Lilli. Consultants are expected to become more tech-enabled, shifting their focus from analytical tasks to activating insights and driving greater value for clients. AI will also enhance diversity in consulting by enabling a broader range of professionals to participate effectively, supported by upskilling opportunities. McKinsey’s experiments with reasoning agents in initiatives like LilliX signal a future where AI capabilities advance further, incorporating logical reasoning and predictive analytics.

To get the latest AI transformation case studies straight to your inbox, subscribe to AI in Action by AIX — your weekly newsletter dedicated to the exploration of AI adoption in business.

Elevate your understanding of AI transformation. Browse AI adoption case studies searchable by company, industry, use case, and technology.

Sources:
What McKinsey learned while creating its generative AI platform

Let’s talk

Whether you’re looking for expert guidance on AI transformation or want to share your AI knowledge with others, our network is the place for you. Let’s work together to build a brighter future powered by AI.