# McKinsey Counts 25,000 AI Agents as Employees: Are You Ready for 1:1?

> McKinsey CEO Bob Sternfels now counts 25,000 AI agents alongside 40,000 human staff. Here is what their 1:1 staffing model means for enterprise operations.

Source: https://www.starbox.work/en/insights/ai-news/2026-03-23-23/ (Starbox AI Insights, updated 2026-03-23)

How many people work at McKinsey?

If you ask CEO Bob Sternfels, his answer is 60,000: 40,000 humans and 25,000 AI agents.

This is not a metaphor or marketing hyperbole. At CES in January 2026, the leader of the world's premier management consultancy officially added AI agents to the firm's formal headcount metric. A McKinsey spokesperson later confirmed the breakdown to Business Insider.

Eighteen months ago, McKinsey ran a few thousand agents. Today, it runs 25,000. Sternfels aims to give every human employee at least one dedicated AI agent partner by the end of 2026, reaching a 1:1 human-to-agent operating ratio.

When the highest-billing knowledge business in the world redefines what constitutes an "employee," every executive should take notice.

## The "25 Squared" Model: Cut Non-Client Roles by 25%, Grow Revenue

Sternfels organizes this shift around what he calls the "25 squared" model. McKinsey is growing client-facing advisory roles by 25% while reducing non-client, back-office roles by 25%. Despite fewer back-office staff, non-client output has increased by 10%.

Historically, professional services scaled linearly: hire more consultants, bill more hours, generate more revenue. McKinsey's new equation breaks that constraint. Headcount expands in front of clients, contracts behind the scenes, and top-line revenue continues to grow. Sternfels describes this as a fundamental paradigm shift for the firm.

Two operational drivers make this model work:

First, McKinsey runs its AI operations through QuantumBlack, an internal team of 1,700 technologists dedicated to AI initiatives. Work led or supported by QuantumBlack now drives 40% of McKinsey's total revenue.

Second, the commercial model is changing. McKinsey is moving away from billing pure consultant hours toward value-sharing agreements tied directly to client business outcomes. When 25,000 digital workers operate around the clock without billing by the hour, selling time is no longer the most profitable strategy.

## EY Pushes Back: Why Headcount Is the Wrong Metric

McKinsey's 25,000-agent milestone sparked debate across the consulting industry.

Steve Newman, Global Head of Engineering at EY, told Business Insider that high agent counts do not necessarily equal business value. Newman noted that EY gets its highest returns from "just a few agents doing the heaviest lifting." Instead of tracking digital headcount, EY evaluates AI impact through operational KPIs: productivity, output quality, and cost reduction.

This disagreement highlights two distinct approaches to enterprise AI deployment:

* **McKinsey's broad pairing model:** Deploy AI partners across the entire organization so every knowledge worker collaborates with software agents daily.
* **EY's targeted deployment model:** Identify a small number of high-leverage bottlenecks and assign specialized agents to resolve them.

Neither model is universally superior. The right choice depends on company size, industry, and organizational maturity. But the industry conversation has shifted. By 2026, leaders no longer debate whether to adopt AI; they debate how to orchestrate it. According to Moltbook-AI, 67% of Fortune 500 companies run at least one AI agent in active production, up from 34% in 2025.

## The Agent Sprawl Trap for Mid-Market Enterprises

McKinsey can deploy 25,000 agents because it has 1,700 technical specialists at QuantumBlack, proprietary corporate knowledge accumulated over decades, and substantial R&D budgets. Most mid-market companies do not.

Chasing high agent counts without structural oversight creates "Shadow AI." When individual employees connect disconnected AI tools to internal data without IT coordination, organizations face immediate risks: data leaks, compliance failures, and contradictory outputs. When agent proliferation outpaces enterprise governance, operational efficiency drops.

McKinsey's own 2025 AI survey mirrors this pattern: while 88% of surveyed organizations use AI in at least one business function, only about one-third have begun scaling it, and under 10% have implemented agentic AI at an operational level. The competitive gap is no longer access to AI tools; it is the presence of an orchestrating system.

## The Starbox AI View: Agents Are Capabilities, Not Headcount

The most instructive takeaway from McKinsey is not the 25,000 figure. It is Sternfels' view of human-machine interaction: agents are no longer passive software tools, but active operational collaborators.

For this collaboration to work, agents cannot operate as 25,000 isolated bots. They must function inside a governed, interconnected architecture.

This principle shapes Starbox AI's enterprise framework. Our three-tier architecture—Organization, BOX (intelligent workspace), and Agents—addresses the challenge of multi-agent coordination. Starbox AI's 300+ specialized agents do not operate as independent tools. They run within a unified knowledge structure, clear permission hierarchies, and structured decision pathways.

McKinsey relies on 1,700 internal specialists to build and govern its digital workforce. Starbox AI provides that same structural rigor to smaller teams. With a four-dimensional knowledge framework that combines pre-trained intelligence, industry benchmarks, proprietary company context, and real-time operational data, lean teams can deploy agents that execute reliably within defined operational boundaries.

EY is right: sheer agent volume does not equal enterprise value. McKinsey is also right: structured human-agent collaboration is the future of knowledge work. Winning teams will deploy agents that run on structured methodologies, validated business context, and systematic governance.

## Key Takeaway for Business Leaders

McKinsey's "25 squared" operational model demonstrates that AI is changing the fundamental mechanics of team design, operational output, and business growth.

If your leadership team had to report total headcount tomorrow, would you include your AI agents? And if you did, do those agents operate under a unified system, or as scattered point solutions?

The answer to that question will define how your organization scales over the next three years.