A year-and-a-half ago, management consulting firm McKinsey had just 3,000 AI agents in its possession, with its 40,000 employees far outnumbering its agentic fleet. But in just 18 months, that number has grown more than 500% to about 20,000 AI agents supporting the company’s work, CEO Bob Sternfels said on Harvard Business Review’s Ideacast. Now, the company is evaluating how well job candidates can work with its AI tool as part of the interview process.
The consulting firm is asking candidates to use its internal AI tool Lilli in a test during its hiring process, according to consulting interview preparation company CaseBasix, which helps candidates solve McKinsey, BCG, and Bain cases. In a blog post, CaseBasix says it gathered information from internal sources who say some candidates would be asked to work with the company’s AI tool as part of a final round AI interview. The Financial Times also reported on McKinsey’s focus on business school students using Lilli, citing people familiar with the matter.
The move comes as the blue-chip company seeks to further implement AI into its operations, pursuing skills that extend beyond the interpersonal and problem-solving traits usually required of a consultant. Companies like McKinsey are looking for candidates who can be AI-ready on day one as the technology becomes essential to job functions.
In his interview with HBR, Sternfels said AI models have developed an expertise in problem-solving, and that the company would be “looking more at liberal arts majors, whom we had deprioritized,” for potential sources of creativity as the firm moves to find creative solutions beyond “logical next steps.” It’s not just McKinsey, other leaders are looking to hire liberal arts graduates like CEO of IT firm Cognizant Technology Solutions Ravi Kumar S, who says he’s recruiting candidates with liberal arts degrees.
Putting AI skills to the test
McKinsey hasn’t shied away from AI in the hiring process. The company encourages AI use in the application process on its career page, saying that candidates can use the technology to refine résumés and practice interview questions. Though it cautions candidates to use the technology responsibly, saying use of the technology during assessments and for generating interview responses, as well as embellishment, is not permitted.
“We welcome those who share our curiosity about AI and its potential,” the company’s career page says.
But the pilot program goes a step further. According to Casebasix, the AI interview may be an additional step in the application process, alongside the case interview and a personal experience interview for candidates in the U.S. And North America.
“In the McKinsey AI Interview, you are expected to prompt the AI, review its output, and apply judgment to produce a clear and structured response,” the Casebasix post said. The post says that McKinsey is looking to test soft skills essential to working at the consulting firm—and for working with the company’s AI—including collaboration and reasoning.
A McKinsey spokesperson did not immediately respond to Coins2Day’s request for comment.
An agentic workforce reshaping the nature of consulting work
Sternfels predicted the company will adopt AI aggressively within the following months. “In another 18 months I think every employee will be enabled by one or more agents,” Sternfels said on HBR’s Ideacast. “We’ll have a workforce that is human and agentic, and we’re going to have to navigate that.”
That change could dramatically shift the work that McKinsey performs. With AI agents making the company’s employees more productive, Sternfels says that the AI adoption could fundamentally change McKinsey’s model.
“We’re migrating away from pure advisory work, away from the fee-for-service model,” Sternfels said. “We’re moving to more of an outcomes-based model, where we identify a joint business case with our clients, and we underwrite the outcome by tying our fees to the impact our work delivers for them”
But the human skills that Sternfels says AI can’t replace: creativity, aspiration, and judgment. “There isn’t truth in AI models; there isn’t judgment,” Sternfel said. “Humans need to impose those parameters.”
