Most leaders are feeling the pressure right now. On one side, there is clear – and often substantial – expectation pressure to take AI seriously. On the other side, an organization built for a different time, with structures, roles, and governance models that have worked well for many years.

It's not surprising that many choose to proceed cautiously. Still, it's worth pausing when some of the world's most traditional and methodical businesses are not only talking about AI as a tool but are beginning to refer to it as part of the workforce. When AI moves from being a tool to a colleague, it's a signal that deserves attention.

A number that changes the conversation

Bob Sternfels, global CEO of McKinsey & Company, recently described his organization in a way that a few years ago would have sounded like science fiction. He refers to the workforce as a combination of around 40,000 people and approximately 25,000 AI agents.

This is not a vision for 2030. It is a realistic description of how the company delivers value today.

When a company that relies on methodology, trust, and precision integrates AI on this scale, it's not an experiment. It's a conscious, strategic choice about how capacity should be understood and managed in a new reality.

When capacity is no longer equal to headcount

For ten years, the number of employees has been the most intuitive way to understand capacity. It is concrete. It is measurable. It is easy to report on.

But this context is about to break down.

At McKinsey, capacity is no longer just the sum of heads, but the sum of people and autonomous systems that actually do the work. Not because humans have become less important, but because the very nature of work has changed.

This enforces a new management principle: leading based on delivery capability, not staffing.

Delivery capacity as a management logic

Delivery capability is about the ability to deliver consistently, predictably, and at the right pace over time – regardless of time zones, working hours, and individual constraints.

Autonomous systems provide a type of capacity that humans alone cannot offer:

• availability around the clock
• no holidays and no sick leave
• consistent quality over time
• immediate scaling with increased volume

People contribute something completely different:

• judgment and ethical reflection
• context understanding and relational intelligence
• responsibility towards customers, employees, and society

When these roles are clearly separated, the organization paradoxically becomes more human. People no longer have to function as machines – because the machines are finally good enough to take over the repetitive tasks.

This is not an IT question

The figures from McKinsey also point to something that often disappears in the AI debate. When AI agents are counted as part of the workforce, it ceases to be an IT issue.

IT can build, secure, and operate the systems. But it is management's responsibility to decide:

• what needs to be delivered, and to whom
• how responsibility and decision-making authority are distributed between humans and machines
• how the organizational chart should look when a significant portion of the workforce never sleeps

As AI becomes part of the actual delivery capability, this will become a matter of organizational design and leadership – not tool selection.

A proof of action, not intention

The most interesting thing about McKinsey's 25,000 AI agents is not the number itself, but what it represents. It is a clear signal of action rather than intention.

They have accepted that technology will have real consequences for how work is organized, measured, and led. It requires leadership that is willing to challenge the management models many of us have learned at BI, NHH, or Harvard – without dismissing them, but by acknowledging their limitations.

An invitation to reflection

This does not mean that all organizations should copy McKinsey. Different businesses have different conditions, cultures, and responsibilities.

But it raises a question that is relevant for every leader who wants to remain relevant in the coming years:

Are we still measuring capacity in number of employees – or are we ready to start measuring actual delivery capability?

It's not a technological question. It's a management question.