In early 2026, McKinsey CEO Bob Sternfels told the audience at the Consumer Electronics Show in Las Vegas something that spread across global media within hours: "When people ask how many people work at McKinsey, I say 60,000 — 40,000 humans and 20,000 agents." He was not talking about robots. He was talking about agentic AI — systems capable of planning, executing, analyzing, and producing outputs autonomously, without step-by-step human instruction.
That number has since grown. By the time of his most recent interview with the Harvard Business Review, the agent count had reached 25,000. The stated goal is to reach parity with McKinsey's human headcount before the end of 2026.
McKinsey's 25,000 Digital Colleagues
This is not a pilot program or a forecast. It is already running. McKinsey's agents — powered largely by an internal platform called Lilli — handle research, document synthesis, comparative analysis, and first drafts of client reports. Functions that, until recently, were the training ground for junior consultants. In a single year, these systems saved the firm an estimated 1.5 million hours of human labor.
McKinsey's business model is being reshaped in the process. The firm is shifting away from pure fee-for-service billing toward outcome-linked payment structures — a change that ripples through hiring practices, client contracts, and the very definition of what a consulting engagement looks like. BCG, PwC, and the major investment banks are moving in the same direction.
Explore how AI is redefining decentralized finance in our dedicated piece on DeFAI: DeFAI — AI and Decentralized Finance
JPMorgan's AI-First Banking Infrastructure
JPMorgan Chase made one of the quietest but most consequential moves of the same period. The bank rolled out its LLM Suite — built on models from OpenAI and Anthropic — to approximately 250,000 employees. Around half of them use it every day. The end goal, in the words of Derek Waldron, the bank's chief analytics officer, is unambiguous: every employee will have a personalized AI assistant, every process will be managed by agents, and every client will have an artificial concierge.
JPMorgan is not testing AI. It is embedding it into its core infrastructure, the same way organizations once wired in electricity.
For context on how institutional AI adoption is intersecting with crypto markets, see our coverage of DeepSeek's performance impact: DeepSeek and Crypto Market Returns
What This Means for Crypto and Web3
The logic driving McKinsey's and JPMorgan's transformation is the same logic now entering DeFi, portfolio management, and algorithmic trading platforms. Agentic AI — systems that plan, execute, and adapt without human supervision — is the engine that will run increasingly complex on-chain protocols without anyone sitting at a screen.
Markets have already started pricing this in. AI models operating autonomously on crypto markets have posted remarkable performance figures, and the structural momentum is not slowing. Binance, for example, is integrating OpenAI into its on-chain infrastructure alongside tokenized pre-IPO assets from SpaceX and OpenAI itself.
See how Binance is building this out: Binance Wallet, Pre-IPO SpaceX and OpenAI Tokenized On-Chain
Here is the tweet that sparked the wider debate on X:
Altman and others need to heavily beef up their personal security.
— Dr Singularity (@Dr_Singularity) April 14, 2026
"A man accused of throwing a lit Molotov cocktail at the home of OpenAI CEO Sam Altman's home last week is being charged with attempted murder, the San Francisco District Attorney said."
"The suspect, Daniel… pic.twitter.com/Ly3yhkiJkI
The Question Nobody Wants to Answer
Sternfels was direct: within 18 months, every McKinsey employee will work alongside at least one AI agent. Roles won't all disappear — but they will change fundamentally. Less routine analysis, more strategic oversight. The problem is that not everyone will be positioned to make that shift.
According to the McKinsey State of AI 2026, 95% of organizations that have invested in AI have not yet generated measurable returns. The divide is no longer between those who use AI and those who don't. It runs between organizations that genuinely integrate it at depth — and those that stop at the chatbot they roll out for board presentations.
The hybrid human-machine workforce has already started. The only question left is which side of that divide you want to be on.
