Good morning. In part 2 of Starting Point’s series on how Boston-area businesses are using artificial intelligence, I’m introducing you to a top consulting firm’s wide-ranging AI tool.
But first, here’s what else is going on:
- The Massachusetts Legislature proposed raising court-appointed lawyers’ pay by $20 an hour over two years to end the work stoppage that has paralyzed the state’s criminal courts. Massachusetts relies heavily on such lawyers to represent poor defendants, and the Legislature’s plan would also hire more state-employed public defenders.
- Brown University reached a deal to restore $50 million in federal research funding and settle the Trump administration’s claims that it failed to protect Jewish students. Unlike Columbia, Brown won’t pay the federal government; it will instead support groups in Rhode Island.
- Former vice president Kamala Harris won’t run for California governor next year, a possible step toward seeking the presidency again in 2028.
Write to us at [email protected]. To subscribe, sign up here.
TODAY’S STARTING POINT
Erik Roth had a problem. In November 2022, the board of McKinsey & Company, the consulting firm where Roth is a senior partner, asked him to make it easier for employees to access and apply the wealth of knowledge the firm had amassed over decades. The status quo was discouraging. McKinsey’s internal digital catalog contained about 65,000 PowerPoint documents. Even using keywords, a search could yield hundreds at a time.
But generative AI technology, which catapulted onto the scene after OpenAI launched ChatGPT that same month, offered a solution. So a few months later, Roth brought the board a demo. After feeding dozens of marketing and sales documents into an OpenAI model, he showed how it could answer the executives’ questions about the documents almost instantaneously.
“They thought it was a joke,” Roth told me. “And that’s where Lilli was born.”
Advertisement
Last week, in part one of a series about how Boston-area businesses are using AI, I took you inside a Charlestown robotics firm that’s using the technology to improve how it trains robots. Today I’ll tell you the story of how McKinsey, the world’s most prestigious consulting firm, built an AI platform to transform its nearly century-old business — and, if it gets its way, lots of other companies’ work, too.
Advertisement
How it works
McKinsey launched Lilli, the AI tool Roth and a team of engineers designed, in July 2023. It’s named for Lillian Dombrowski, who in 1945 became the first professional woman the firm hired. Technically speaking, Lilli is an “orchestration layer,” working across 11 different AI models to execute tasks. McKinsey wouldn’t let me publish photos of Lilli (it has three patents on the tool already and several more pending). But in a conference room in McKinsey’s Boston offices, high above the Financial District, I saw it in action.
Opened in a browser window, Lilli looks spare and unassuming. Then Roth keyed in a very consultant-y sounding task: Break down the electric-vehicle battery industry value chain into steps. After thinking for a few moments, denoted by a circular loading icon, Lilli generated a formatted list that spanned mining raw materials to recycling spent batteries.
The goal is to draw on as much relevant knowledge as possible. Lilli pulls from McKinsey’s PowerPoints, expert interviews, and other internal sources, plus public data about thousands of companies and industries. Its EV list featured numbered footnotes that resembled a Wikipedia page; blue citations indicated an internal McKinsey source, orange an external one. (In small text, Lilli warns that it “may produce inaccurate or biased information.” Consultants verify its output and provide feedback using thumbs-up or thumbs-down icons.)
Two years after Lilli debuted, two-thirds of employees across McKinsey’s scores of offices around the world use it monthly; more than forty percent use it weekly. “It’s become a second screen for most of our people where it’s up all day long as a problem-solving partner,” Roth said.
Advertisement
Welcome to the revolution
That might not sound particularly transformative. But what makes Lilli different is McKinsey’s effort to merge it with the company’s other functions. Lilli can create shareable PowerPoint slides of content it’s generated, rewrite text to fit the firm’s house style, and suggest who in the firm a consultant should ask for more information.
Rob Levin, a senior partner in McKinsey’s Boston office, sees AI as a transformational tool for businesses that adopt it widely. “This is the future of competitive advantage,” he said. McKinsey shows off Lilli to clients that might like to implement their own version.
Lilli also keeps evolving. The firm recently launched AI agents — “virtual helpers,” in Roth’s shorthand — that can execute tasks without additional prompts. For example, a consultant could have an agent generate a profile of a company and a memo appealing to its needs — or even to specific executives. McKinsey is also rolling out a voice-enabled feature; Lilli’s own voice, which I heard during my visit, is peppy and eager.
As Lilli could probably tell you, revolutions can be bloody, and some of AI’s drawbacks are becoming clearer. People who talk to chatbots can have their delusions magnified. Students can shortcut learning. In a recent Greater Boston Chamber of Commerce survey about how Massachusetts companies are using AI, some reported concerns about data security, employee resistance to AI adoption, and even layoffs. Levin imagines that prompting AI effectively will become a core skill for many workers. And if AI helps a company grow, it might hire more.
For now, Lilli is primarily saving time. Before it, Roth said, a consultant might’ve needed a day or more to figure out which EV battery experts to call and which information to use.
Advertisement
Could that mean less time spent working? Roth suggested it’s already happening in limited ways. “We’re saving the average consultant over one full day per week,” he said. “If they’re getting in that extra gym class during the day, or whatever they might be doing to bring back some balance, it’s not a terrible thing.”
But a four-day workweek? Now that would be revolutionary.
