Is McKinsey’s move to count 25,000 AI agents as “employees” a blueprint for how labor, value, and headcount accounting are about to shift?

The real story behind “25,000 AI employees”

Saying McKinsey now has 60,000 employees, of which 25,000 are AI agents, is a deliberate act of reframing, not a slip of the tongue. In interviews and conference appearances over the past few weeks, CEO Bob Sternfels has repeated some version of the same math: roughly 40,000 humans, nearly 25,000 agents, with a goal of giving every employee at least one agent within the next 18 months.

Eighteen months ago, those agents numbered only in the low thousands, which means McKinsey has scaled an entirely new category of labor faster than most enterprises can roll out a new HCM module.

What makes this moment different is that McKinsey isn’t describing these systems as tools; they are positioning them as co‑workers that sit on the org chart next to humans. The firm claims the agents have already saved 1.5 million hours in a year and generated millions of charts, freeing consultants to focus on client‑facing strategy and decision work.

But wait, haven't I read about other consulting firms just copying and pasting from ChatGPT and charing millions only to be found out?

Nonetheless, leaders in HR and talent can no longer dodge the question is no longer whether AI belongs in the business. It is quickly becoming how how you account for, govern, and deploy this new layer of digital labor alongside your people, just as you would any other workforce segment.

I can't wait to so how the BLS will track this.

From headcount to labor portfolio

Most organizations still treat AI as a line item in IT budgets rather than a component of the workforce portfolio. Procurement negotiates licenses, security signs off, and a few enthusiastic teams experiment at the edges, but the systems are not mapped the way finance maps FTEs, contractors, or BPO relationships. McKinsey’s move to publicly describe “40,000 humans and 25,000 agents” forces a different mental model: AI as labor capacity that can be counted, costed, and strategically positioned. When a firm of that size and influence starts talking this way, clients and competitors will be pushed to follow.

This is where HR and Talent Acquisition could fall behind. Many HR teams remain locked in systems that can barely reconcile internal and external talent, let alone a third category of autonomous digital workers. Yet the work has already changed.

Surveys from sources such as McKinsey show that a growing majority of employees use generative AI at least occasionally and that managers see high success rates when they recommend gen‑AI tools to resolve work challenges. If your headcount reports do not reflect this blended reality, your workforce planning will always seem misaligned with how work really gets done.

In earlier articles on topics like Sprint Recruiting and agile talent models, this same gap between “what the system thinks is happening” and “what work actually looks like” proved to be the hidden constraint that kept HR reactive instead of strategic. Or at least that was my experience.

Designing the human–agent operating model

McKinsey’s numbers grab attention, but the more important question is how they are re‑drawing workflows so humans and agents actually complement each other. Early descriptions from the firm suggest a pattern. AI agents handle research, document drafting, and synthesis, while human consultants focus on sense‑making, client engagement, and higher‑order problem framing. In other words, the operating model is shifting from “consultant does end‑to‑end task” to “consultant orchestrates a mesh of agents and humans,” much like a project manager coordinates across internal teams and vendors. That shift requires explicit role design, performance expectations, and new intake rituals that most organizations have not yet defined.

For HR leaders, this is where the job shifts from AI awareness to AI architecture. A future‑ready people function will sketch out which parts of each role are agent‑addressable, which must remain distinctly human, and how to re‑bundle the remaining work into roles that feel meaningful rather than hollowed out.

This echoes the move many companies made when they first centralized recruiting into COEs or introduced shared services. The change coming will be how the shared services layer is populated by software agents that run 24/7 and never attend a town hall. Intake conversations will need to add a new question: “What agents are assigned to this role, and how will we measure the handoff between human and machine?”

HR’s new accountability: governing digital coworkers

Treating AI agents as “employees” creates both symbolic clarity and governance headaches.

On the positive side, it makes accountability for outcomes more explicit. If an agent is effectively logged as part of the workforce, then issues like bias, data leakage, and performance drift cannot be shrugged off as IT problems. They become core HR and risk topics. McKinsey’s own research has warned that AI adoption without strong guardrails can amplify existing inequities and expose firms to reputational damage, even as it boosts productivity. Once you accept the premise that agents sit inside the labor portfolio, it follows that they need the same rigor around selection, onboarding, performance review, and “role retirement” that you apply to humans, even if the mechanisms are technical rather than contractual.

Honestly, we should be operating like this today with any process involving AI, agent or no agent.

Yet HR rarely owns the design of AI governance today. In most enterprises, model selection and monitoring live in data or engineering teams, while ethics and legal review operate in parallel. This fragmentation mirrors the “siloed talent stack” problem I have written about elsewhere, where recruiting, L&D, and workforce planning pull in different directions.

A more coherent path would position HR as the orchestrator of an integrated labor governance model that covers humans, contractors, and agents under a single set of principles. That means defining minimum acceptable standards for explainability, documenting where agents sit in critical workflows, and partnering with risk functions to decide when a human must remain in the loop. The resulting operating model looks less like classic HR policy and more like product governance, with HR leaders behaving as portfolio managers for all forms of labor, not just human employees.

What talent leaders should do now

If you run talent or HR, the McKinsey headline is an early indicator of the questions your CEO will soon be asking you. Boards will want to know how many agents are in the business today, how they affect productivity, and where the labor mix should be in three years to stay competitive. They will also ask where new forms of value are emerging and whether your people function is organizing talent around those opportunities or clinging to legacy org charts.

McKinsey is already signaling that AI is allowing it to grow client‑facing roles by roughly a quarter while trimming non‑client‑facing work at a similar rate, effectively redeploying capacity toward revenue and impact. That is a people story, not a technology one, and it belongs squarely in HR’s remit.

For practical next steps, talent leaders can start by:

Taking an inventorying where agents already exist, even informally, and mapping them to specific workflows in recruiting, learning, and workforce planning.

From there, the work is to build a simple portfolio view: human FTEs, contractors, BPO, and AI agents, all expressed in a common unit of capacity so that scenario planning becomes possible. This is the same discipline that underpins approaches like Sprint Recruiting, where work is broken into units and flows are optimized across teams instead of around individual requisitions.

Finally, leaders should pressure‑test HR’s own operating model: if a consulting firm can scale to 25,000 agents in under two years, what would it take for your people function to operate as a true talent portfolio manager, orchestrating humans and digital coworkers with the same speed and clarity?

Are we really staring down the barrel of a change in employee classification? Not yet. I think this is just the latest move of a CEO trying to look hip and in touch for the board and showing off to competitors. Though, I would be remiss if I did not warn you that it is in the future and you SHOULD begin thinking how to tackle this now. Better to be prepared than to have to try to figure all of this out in the midst of the change.

Good news is, the Department of Labor is a bit tied up now with trying to figure out if the numbers published for the market are legit. And they're counting employees the "old way." So I think you have time, just don't be the last to the dance when the music does finally begin to play.

If you liked this...

• Check out some of my recent posts on parallel topics:

• AI Jobs Clarity Act Breakdown: https://youtu.be/tH_OhrxPrY4

• Trade Jobs are Making a Comeback: https://youtu.be/y3BR5R6SI_o

• HR Trends-Buzzworthy or BS with Matt Charney: https://youtu.be/d_tfr3gJWKs

• 🎧 Available on Spotify, Apple Podcasts, and all major platforms: https://www.purpleacornnetwork.com/podcasts/the-human-capitalist