TL;DR:
- McKinsey introduces its own generative AI tool, Lilli.
- Lilli is designed to provide insights, data, and expert recommendations.
- Built by McKinsey’s “ClienTech” team, Lilli is named after the firm’s first female hire.
- Lilli’s beta version launched in June 2023, set for a full rollout soon.
- Already used by 7,000 employees, Lilli significantly accelerates research and planning.
- Lilli’s interface resembles public text-to-text AI tools with unique features.
- Two tabs: “GenAI Chat” sources data from generalized model, “Client Capabilities” from McKinsey’s corpus.
- Lilli attributes sources transparently, setting a precedent for accountability.
- Lilli finds utility in numerous consulting tasks, from research to project planning.
- Demonstrated versatility in identifying experts, predicting market trends, and formulating plans.
- McKinsey prioritizes quality over speed, aims to enhance response times.
- Future plans include integrating client data for deeper collaboration.
- Lilli combines established large language models with a secure layer for optimal performance.
- Potential white-labeling and external-facing applications are under consideration.
Main AI News:
In a groundbreaking move, consulting powerhouse McKinsey and Company has harnessed the potential of generative AI, introducing its own transformative tool named Lilli. Acknowledged for its swift integration of AI solutions, the firm announced earlier this year that almost half of its 30,000-strong workforce had embraced generative AI technology. This strategic pivot has now culminated in the launch of Lilli, an innovative chat application meticulously crafted by McKinsey’s forward-looking “ClienTech” unit, under the guidance of Chief Technology Officer (CTO) Jacky Wright. Lilli serves as an indispensable resource, furnishing employees with invaluable information, insights, data, plans, and astute recommendations for the most relevant in-house experts, all derived from a colossal repository comprising over 100,000 documents and interview transcripts.
Erik Roth, a seasoned senior partner at McKinsey who spearheaded the product’s development, encapsulated Lilli’s essence in a recent interview with VentureBeat, stating, “Imagine posing a question to the entirety of McKinsey’s reservoir of knowledge and having an AI furnish an insightful response – that’s precisely the role Lilli fulfills.” Aptly named after Lillian Dombrowski, the trailblazing woman who became McKinsey’s inaugural female hire for a professional services role in 1945, Lilli’s beta version was unveiled in June 2023, with a full-scale rollout planned for the upcoming autumn.
Lilli’s Impact: Enhanced Productivity and Empowerment
Roth and his dynamic team at McKinsey divulged that Lilli had been actively utilized by around 7,000 employees, initially as a “minimum viable product” (MVP). Its transformative influence has already been evident, drastically trimming the time allocated to research and planning endeavors from weeks to mere hours and in select instances, even minutes. Roth elaborated, “In just the past fortnight, Lilli has adeptly addressed an impressive 50,000 queries. An impressive 66 percent of users find themselves returning to it multiple times each week.”
Decoding Lilli’s Operational Dynamics
Navigating Lilli’s interface will feel familiar to those acquainted with other publicly-accessible text-to-text generative AI tools, such as OpenAI’s ChatGPT and Anthropic’s Claude 2. Positioned at the base of its primary window, Lilli features a text entry box, enabling users to input queries, prompts, and searches. The ensuing responses unfold chronologically in a conversational format, with user inputs juxtaposed with Lilli’s ensuing elucidations.
However, distinctive attributes leap forth in terms of additional functionality: Lilli boasts an expandable left-hand sidebar housing saved prompts. Users can conveniently copy, modify, and apply these prompts according to their preferences. Roth indicated that prompt categories are slated for integration on the platform in the near future.
Two Transformative Tabs: GenAI Chat and Client Capabilities
Lilli’s interface encapsulates two pivotal tabs for users to effortlessly toggle between. The first, aptly labeled “GenAI Chat,” taps into a generalized large language model (LLM) backend, sourcing relevant data for responses. The second tab, “Client Capabilities,” extracts responses from McKinsey’s extensive repository housing over 100,000 documents, transcripts, and presentations.
Innovatively, Lilli’s Responses Offer Transparent Attribution
Unlike several LLMs that often omit sourcing references, Lilli bucks this trend by incorporating a dedicated “Sources” section beneath each response. This unique feature empowers users with links and page numbers, delivering a comprehensive attribution of the information’s origin. As Roth enthusiastically shared, “We are committed to full attribution. Our clients appreciate this approach.”
Lilli’s Multifaceted Utility
Given its vast access to information, the possibilities for leveraging McKinsey’s Lilli AI are wide-ranging. Roth envisions Lilli as an indispensable ally for McKinsey consultants at every juncture of their client interactions. From procuring initial insights into a client’s industry landscape, competitors, or analogous firms, to sculpting actionable plans for project implementation, Lilli proves to be an indispensable partner.
A Glance into Lilli’s Versatility
VentureBeat’s immersive demonstration of Lilli showcased its versatility to full effect. The AI adeptly identified internal McKinsey experts qualified to discuss a prominent e-commerce retailer, furnished an outlook on clean energy prospects in the U.S. for the next decade, and meticulously devised a blueprint for erecting a new energy plant within a succinct 10-week timeframe. Notably, throughout these multifaceted responses, Lilli unequivocally referenced its information sources, ensuring transparency and accountability.
Quality Over Speed: McKinsey’s Uncompromising Approach
While Lilli’s response times were occasionally marginally slower than the leading commercial LLMs, Roth emphasized McKinsey’s unwavering commitment to quality. The firm continually refines Lilli’s swiftness while safeguarding the integrity and depth of information provided.
Pioneering Future Avenues: Client Data Integration
Roth unveiled an ongoing exploration within McKinsey to facilitate secure, private analysis of client information and documentation on McKinsey servers. Although this transformative feature remains in development, it holds the promise of revolutionizing data synthesis and exploration, fostering deeper collaboration between McKinsey and its clients.
Underpinning Technology: The Unique Stack
Lilli draws upon established large language models, including those developed by McKinsey partner Cohere and OpenAI on the Microsoft Azure platform. However, Lilli stands apart as an innovative secure layer interposed between users and underlying data. Roth elucidated, “Lilli embodies its own stack, an amalgamation of technologies that bridges the gap between the corpus and the LLMs. It amalgamates deep learning capabilities and trainable modules to create a comprehensive stack.”
Embracing Evolution: A Promising Trajectory
Roth highlighted McKinsey’s adaptability in its AI approach, remaining “LLM agnostic” and consistently exploring new AI models to ascertain optimal utility. As the firm contemplates expanding Lilli’s integration across its entire workforce, Roth hinted at the potential for white-labeling Lilli or even extending it as an external-facing tool for McKinsey clients and beyond. He asserted, “All options are on the table. In my view, every organization would benefit from a version of Lilli.”
Conclusion:
McKinsey’s introduction of Lilli marks a transformative step in the consulting industry. The proprietary AI tool streamlines information retrieval and expert recommendations, increasing efficiency and accuracy in consulting processes. With transparent sourcing and versatile functionality, Lilli sets a new standard for AI-powered insights. As the market witnesses McKinsey’s groundbreaking approach, the demand for AI tools tailored to specific industry needs is expected to rise, prompting other organizations to explore similar innovative solutions.
