McKinsey unleashes internal generative AI tool for employees

23 August 2023 3 min. read

Following the likes of a number of its contemporaries, global management consulting powerhouse McKinsey & Company has now unleashed its own, internally designed generative AI assistant: Lilli.

McKinsey & Company has become the latest consultancy to unveil its own, in-house generative AI tool, dubbed ‘Lilli’ – named for Lillian Dombrowski, the firm’s first professional female hire.

While quick to embrace other tools – the firm recently stated that almost half of its 40,000 global headcount was already using the technology – Lilli has been developed by McKinsey’s QuantumBlack division and specifically tailored for employee use.

McKinsey unleashes internal generative AI tool for employees

According to Sydney-based QuantumBlack partner Helen Mayhew, Lilli has fielded 50,000 enquiries in the past two weeks alone. “It’s a platform that provides search and synthesis of the firm’s vast stores of knowledge to bring our best insights, quickly and efficiently, to clients,” she stated on LinkedIn. “It allows our teams to spend less time researching and trawling document repositories, and more time problem solving and implementing real-world solutions.”

Those repositories include more than 100,000 documents, case studies and playbooks across 40 curated knowledge systems – described as the ‘life-force’ of the firm. Now, users can scan McKinsey’s entire intellectual property archive through a simple question prompt and quickly identify the most relevant pieces of content, summarise its key points, and locate expert colleagues in a matter of minutes, saving hours and potentially even weeks of research legwork.

“Lilli aggregates our knowledge and capabilities in the one place for the first time and will allow us to spend more time with clients activating those insights and recommendations to maximise the value we can create,” said McKinsey senior partner Erik Roth, who led Lilli’s development. “An engagement team will be able to spend more time on problem solving, coaching, building capabilities, and helping clients achieve the performance they aspire to achieve.”

In a discussion with Venturebeat, Roth identified several key points of difference between Lilli and most publicly-available models, including attribution and links to the sources the AI draws its responses from, even down to page numbers. Users can also toggle between its primary interface and a second, more generalised open source chat tool to compare the firm’s internal knowledge database with information that is in the public domain.

Having been in beta mode since June and available to approximately 7,000 employees (with two thirds returning to it multiple times per week, according to Roth), Lilli will now be rolled to the remainder of McKinsey’s global workforce over the course of the year, with staff to be provided with training in areas such as ‘prompt engineering’ and content validation. McKinsey will also help clients in building their own tools by sharing its own lessons learned.

As for yet another AI assistant being given a traditionally female name? In this instance, there is solid reasoning behind the choice beyond the stereotypes. Dombrowski was employed as an accountant at McKinsey in 1945 before later serving as corporate secretary over a thirty-five year career. Notably, she was also the firm’s first archivist; “Her long association with leaders and projects meant she knew what records to look for and exactly where to find them.”