McKinsey calls AI 'existential' as consulting's billable-hours model faces disruption

Summary

McKinsey is confronting a structural threat from AI that cuts directly against its billable-hours model. A McKinsey partner flagged the internal contradiction: adopting too much AI will hurt the firm's bottom line. Competitors like law firms face the same pressure but avoid stating it publicly, since appearing anti-tech carries reputational risk. The math is straightforward. If AI agents handle the work that junior consultants spend 80-hour weeks grinding through, the firm loses billable hours and revenue.

McKinsey has already begun to address the gap. The firm reduced headcount from 45,000 in 2023 to 40,000 through layoffs and attrition, while deploying roughly 12,000 AI agents. The comparison is imprecise because agents spin up and down without representing persistent staff, but the direction is unmistakable.

Talent identification and retention

The grueling 80-hour weeks serve a purpose beyond hazing. They function as an extended interview for identifying grinders, creative thinkers, and people who can build relationships with Fortune 500 executives. One partner described an analyst who flew to a developing nation, sat with top government officials, modeled a take-private deal, and returned with a letter of recommendation from the prime minister that managed to be both glowing and worded like a threat to Harvard Business School. That impact is not something AI agents can replicate.

McKinsey's risk is in talent identification and retention. If you collapse the on-ramp into pure evaluation without the crucible of actual high-stakes work, you lose the signal for who thrives under pressure. You also lose the mentorship pipeline. Not every top performer stays. Some burn out, some leave to start companies, some arrive from tech or entrepreneurship later. The grinding period is where the firm learns who can lead.

A second threat runs through client relationships. When a high-performing analyst builds a relationship with a Fortune 500 CFO or COO, they could theoretically leave and start their own firm, taking the client with them. Historically, they would need to hire and train junior consultants. But if ChatGPT Pro gives them 40 virtual consultants on day one, or 12,000 if they can deploy McKinsey's own agents, the cost of defection drops sharply. The barrier to spinning up a boutique consulting shop becomes much lower.

What consulting sells

Consulting's value proposition has long rested on an ambiguous foundation. Do clients hire McKinsey for strategy insight, or for cover—a way to outsource critical decisions so executives can say "McKinsey told us to do this"? Both occur. That defensibility becomes harder to maintain with AI. Once the grunt work falls away—data synthesis, market analysis, deck-building—the firm's competitive edge collapses into pure human judgment and relationship capital. That is a smaller, higher-touch business. It resembles a partnership of senior dealmakers with access to best-in-class tooling, whether Harvey for legal or OpenAI for everything else, rather than a 40,000-person factory.

McKinsey's advantage is its private structure, which allows longer-term bets. Accenture is public and roughly 20 times larger at 791,000 employees. Its stock has dropped 35% in six months, bookings are slowing, and government work, a major revenue line, has been disrupted by spending cuts. McKinsey has more breathing room, but the clock is ticking.