McKinsey & Company launches Ask McKinsey – Gen AI chatbot for research discovery
Global management consultancy McKinsey & Company has released a new generative AI chatbot for easy exploration of the firm’s published research linked to their original source materials.
Dubbed ‘Ask McKinsey’, the chatbot was built in the space of five months on the back of McKinsey & Company’s own internal AI assistant Lilli, which was rolled out in 2023, with additional moderation and risk measures in place.
Described as a professional services sector first, clients and other readers of McKinsey’s research material can now type in a specific question and receive a direct outline in response drawn from its database with links to the relevant papers and articles.
“Ask McKinsey isn’t just a new feature – it’s an example of how we’re reimagining the way knowledge is delivered,” the firm said. “As search behaviour shifts toward conversational AI, we’re offering a tool that’s both cutting-edge and grounded in our trusted research, reflecting our goal to deliver meaningful differentiation, and not just better sameness, to our audiences.”
In addition to risk and safety, one area of concern for the firm was to develop a chatbot which would be consistent with its status, long considered the most prestigious management consultancy in the world, with around 50 publishing and technology team members working together to build evaluation data-sets, test tone, and fine-tune responses.
Speaking on the roll-out, Marianne Blum, Mckinsey’s US-based director of external engagement, said: “We ensured that the bot doesn’t just provide accurate information – it does so in a way that is consistent with the clarity, authority, and accessibility readers expect from McKinsey insights.”
Functional advantage
A test-run by Consultancy Australia of Ask McKinsey confirms responses which are succinct and delivered in plain language, and accurate as far as the firm’s original research is, but there are limitations for now, naturally in terms of being restricted to research undertaken by the firm. There is also as yet a function to search within a specific paper, which can sometimes run upwards of 100 pages.
For example, when questioning the bot on ‘infrastructure trends in Australia’, the user is provided an outline of various statistics, challenges, and a brief conclusion with recommendations, linked to a paper from 2019 – with the firm presumably not having broached the subject locally in more than a half a decade during which time there have been some major market developments and disruptions.
The feature does provide the disclaimer in its conclusion - “This analysis is based solely on insights published by McKinsey,” – but taking a more recent piece of research, on the impact of generative AI on the Australian workforce, and the functional limitations of the tool extend to the inability to direct to specific data within that text, such that it acts more like an advanced database index.
For clarity; should a user type, ‘How many Australians will lose their jobs due to AI?’, the bot provides the answer, “Up to 1.3 million workers may need to transition out of their current roles into new occupations by 2030” as part of its summary, linking to McKinsey’s paper from last year. Clicking on that link however simply takes the user to that 62-page paper, not the specific section.
In launching its new AI tool, McKinsey states, ‘Readers expect trusted guidance – not endless scrolling”, but it’s difficult to ascertain any fresh advantages over a standard web-browser search and in-text Control+F function using key words to hopefully find the relevant passage, and then externally Google the citations to then evaluate the credibility of the data source and conclusions.
It’s unclear if Ask McKinsey is still in beta mode (it still has such a tag, although the firm says it’s now live across all of its online insights), but the consultancy has flagged a number of potential future upgrades, including personalised suggestions (users have to register or access via an account such as LinkedIn or Google), interactive exhibits and charts, and, notably, deeper integration.
