The Great Management Rewrite: Moving from ‘Span of Control’ to ‘Span of Intelligence’

CULTURE & CODE

How many people can one manager effectively manage?

For more than a century, that question has helped determine how organizations are designed. It influences spans of control, management layers, reporting relationships, team structures, and ultimately the boxes and lines we draw on organizational charts.

But AI introduces a very different question:

What happens when people are no longer the only intelligence a manager is expected to manage?

Imagine a manager leading eight employees while simultaneously directing a Research Agent, Analytics Agent, Customer Insight Agent, Scheduling Agent, Content Agent, and several specialized agents embedded inside different workflows. Some gather information. Others analyze it. Some recommend actions. Others execute routine work continuously in the background.

The manager still manages people — but increasingly also orchestrates digital capability.

This is no longer a distant thought experiment. Microsoft’s 2025 Work Trend Index describes the emergence of the “Agent Boss” — workers who build, delegate to, and manage AI agents as part of their jobs. Microsoft reported that 81 percent of leaders expected agents to be moderately or extensively integrated into their companies’ AI strategies within 12 to 18 months, while 36 percent expected managing agents to become part of employees’ jobs within five years.

Those numbers suggest something much bigger than another productivity tool entering the workplace.

For more than a century, management has been designed around a workforce composed almost entirely of human beings.

What happens when that assumption is no longer true?

Welcome to the Great Management Rewrite.

 

Management Was Designed Around Human Constraints

Much of modern management exists because human capability has natural limits.
People have finite time and attention. Information takes time to gather. Coordination requires conversations and meetings. Managers can review only so much work, absorb only so much information, and maintain only so many meaningful relationships before effectiveness begins to decline.

Organizations adapted to those constraints by creating hierarchy.

Information moved upward. Decisions moved downward. Managers coordinated the work in between. As organizations became larger and more complex, additional managerial layers helped process that complexity.

AI begins to alter some of those assumptions.

Microsoft’s concept of the Frontier Firm describes organizations increasingly built around human-agent teams, where agents expand the capacity available to individuals and groups. McKinsey’s Superagency in the Workplace similarly argues that AI’s larger opportunity comes from amplifying human agency when organizations redesign work around the technology rather than simply layering AI onto existing processes.

That distinction is critical.

If AI simply helps managers write emails faster, summarize meetings, or prepare presentations, management becomes more efficient.

But if AI can continuously gather information, monitor workflows, surface anomalies, synthesize data, generate options, coordinate routine activities, and execute certain processes autonomously, something more fundamental happens –

The architecture of management itself begins to change.

The manager no longer needs to be the primary processor of information flowing through the team. Instead, managerial value can move toward something technology finds much harder to provide: judgment, context, prioritization, coaching, motivation, trust, accountability, and meaning.

Perhaps AI will not eliminate management –

It may eliminate some of the work that has prevented managers from actually managing.

Delegation Is About to Be Rewritten

This becomes particularly visible when we think about delegation.

Traditionally, delegation has been human-to-human. Managers decide what work needs to be done, who should do it, how much authority to provide, and when they should intervene.

Now introduce AI agents into that equation.

The manager faces an entirely new set of choices. Which work belongs with a person? Which belongs with an agent? Which should involve both? How much autonomy should the machine receive? When should human review be mandatory? When is machine speed more valuable than human discretion — and when is human judgment worth the additional time?

Anthropic’s Economic Index is useful here because its research into actual AI usage distinguishes between automation (where AI performs more of the task) and augmentation (where AI and humans collaborate). This reinforces an important principle: the impact of AI does not occur uniformly across entire jobs. Different tasks lend themselves to different combinations of human and artificial intelligence.

Delegation therefore becomes more sophisticated than deciding: “Who should do this?”

Tomorrow’s manager must increasingly ask: “What combination of human and artificial intelligence should do this best?”

That is not merely AI literacy.

It is a new management capability.

 

From Span of Control to Span of Intelligence

And this leads to what I believe could become one of the more consequential changes in organization design.

Management has traditionally relied on the concept of span of control: how many people can one manager effectively supervise?

But what happens when the productive capability available to a manager is no longer represented by headcount alone?

Perhaps we need another concept – Span of Intelligence.

I define this as the total human and artificial capability that one leader can effectively orchestrate toward an outcome.

Imagine two managers, each with ten employees.

The first manages ten people conventionally.

The second manages the same ten people, but the team also operates with AI agents supporting research, analysis, administration, customer intelligence, knowledge retrieval, workflow monitoring, and routine execution.

On an organization chart, both teams appear identical.

In terms of effective capability, they may be completely different.

This is why headcount alone may eventually become an incomplete measure of organizational capacity.

A team of ten people supported by twenty specialized agents may possess more productive capability than a traditional department of fifty. A small management team equipped with sophisticated decision support may coordinate complexity that once required several organizational layers.

AI is not yet expected to automatically produce wider spans of control or flatter organizations, so it would be premature to present that outcome as fact. What the research does indicate is that AI is reshaping tasks, skills, and human-machine collaboration.

The World Economic Forum’s Future of Jobs Report 2025, for example, identifies AI and information-processing technologies among the major forces expected to transform businesses through 2030, while continuing to emphasize human capabilities including analytical thinking, resilience, leadership, social influence, and lifelong learning.

And then, there is an interesting paradox emerging here –

As artificial capability expands, the value of distinctly human managerial capability may increase with it.

The manager’s scarce resource stops being just “access to information.”

It becomes “the judgment required to know what to do with all that intelligence.”

 

Managing Two Different Kinds of Performance

That creates another problem organizations have barely begun to address.

Performance management was designed for people.

We establish objectives, define KPIs, review outcomes, diagnose capability gaps, provide coaching, recognize contribution, and address underperformance. But what happens when a meaningful portion of the team’s productive capacity is artificial?

A manager of a hybrid team must eventually understand not only whether the people are performing, but whether the system of people and agents is performing.

Is the agent accurate? How frequently does its output require correction? Is it improving decision quality or merely increasing output? Is it being deployed against the right tasks? Does the value created justify its cost? Is the human-agent workflow actually better than the process it replaced?

These questions connect with something I’ve previously called Return on Intelligence: the value organizations create from the intelligence — whether human and artificial — which they deploy.

That concept becomes particularly relevant at team level.

The objective isn’t to maximize human activity. Nor is it to maximize AI usage.

It is to maximize the value created by the combination.

Tomorrow’s manager may therefore manage two performance systems simultaneously –

The development and contribution of human talent, and the effectiveness and economics of digital capability.

Eventually, perhaps we should stop thinking of them as two systems at all.

The more useful unit of performance may become the hybrid team.

You Can Delegate Execution, But You Cannot Delegate Accountability.

There is another boundary that becomes more important as agents become more capable.

Autonomy is not accountability.

A manager may delegate research to an agent. The agent may analyze thousands of records, prepare recommendations, draft communications, or execute parts of a workflow with minimal intervention. But someone still has to decide what outcome matters, what boundaries apply, what risks are acceptable, and whether the result should ultimately be acted upon.

McKinsey’s Superagency research emphasizes that capturing AI’s potential requires more than giving people access to technology; organizations need leadership, capability building, and redesigned ways of working. Deloitte’s Global Human Capital Trends similarly places human performance and organizational choices at the center of technological transformation rather than treating technology as an isolated solution.

This creates another reversal in managerial value.

When information is scarce, managers who possess information have power.

When information becomes abundant, discernment becomes more valuable.

When analysis takes days, producing analysis is valuable.

When analysis takes seconds, knowing which analysis matters becomes more valuable.

And when execution can increasingly be delegated, deciding what should be executed, why, and with what consequences becomes more important – the manager of the future therefore becomes less of a task supervisor and more of a judgment architect.

AI can expand managerial intelligence.

But leadership must remain accountable for how that intelligence is used.

 

The Hardest Part of the Digital Team May Be the Human Team

For all the discussion about agents, there is an irony here.

The hardest part of managing a digital workforce may have very little to do with the digital workforce. It may be managing the humans working beside it.

Imagine being told that your newest teammates can research continuously, analyze enormous datasets, generate work in seconds, and perform certain activities at a fraction of the time previously required. Employees will naturally ask questions that are far more personal than technological: What happens to my job? Which parts of my expertise still matter? How will I progress? What should I learn? And if the machine can now perform work that once demonstrated my value, where does my value move next?

Those questions cannot be answered with another AI training course.

They require leadership.

The World Economic Forum’s findings are instructive here. Even as technological capabilities grow in importance, human capabilities such as analytical thinking, creative thinking, resilience, leadership, collaboration, and lifelong learning remain central to the future workforce.

The manager’s responsibility therefore isn’t simply to make employees better at using AI –

It is to help employees move toward the work where being human creates disproportionate value.

That means helping people surrender tasks where machines possess overwhelming advantages while developing the judgment, creativity, relationships, empathy, context, leadership, and problem-solving abilities that become more valuable precisely because machines are becoming more capable.

AI adoption therefore becomes talent development.

And digital transformation becomes a deeply human management responsibility.

HR Must Rewrite Management, Before AI Does

This is where HR has an enormous role to play.

For decades, management development has been built around a familiar set of competencies: communication, delegation, coaching, performance management, decision-making, conflict resolution, and leadership.

None of those disappear. But, they may no longer be enough.

Tomorrow’s managers may also need to know how to frame problems for AI, allocate work between humans and agents, determine appropriate levels of autonomy, challenge machine-generated recommendations, verify outputs, understand the economics of digital work, and design workflows where human and artificial intelligence complement one another.

That means organizations may eventually need to rethink far more than AI training.

Management competency models may need to change.

Leadership development may need to change.

Performance management may need to change.

Job architecture may need to change.

Succession planning may need to change.

Even the criteria by which organizations identify high-potential leaders may need to change.

As Microsoft calls the emerging worker who directs agents an “agent boss,” I think the leadership implication goes even further.

We may be witnessing the evolution of the People Manager into the Intelligence Orchestrator — a leader whose job is neither to extract maximum productivity from people nor maximum utilization from AI, but to create something greater from both.

HR should probably start designing for that manager now.

Because technology is already designing the job.

 

The Great Management Rewrite

Management has survived every major technological revolution because management itself has continually evolved.

Industrialization changed it. Mass production changed it. Computers changed it. The internet changed it. Globalization changed it.

AI will change it again.

But perhaps the most important consequence will not be that managers suddenly have digital colleagues alongside human ones. It may be that access to extraordinary amounts of artificial capability forces us to rediscover what management was supposed to be about all along.

Not moving information from one layer to another. Not monitoring activity for activity’s sake. Not spending entire days coordinating meetings about work rather than creating the conditions for great work.

Management at its best has always been about setting direction, exercising judgment, developing people, building trust, making difficult choices, allocating scarce resources, creating meaning, and accepting responsibility for outcomes.

AI doesn’t make those responsibilities obsolete. It strips away some of the work surrounding them — and potentially makes them more important.

For more than a century, organization designers have asked:

How many people can one manager effectively manage?

The AI era gives us a much more interesting question:

How much intelligence can one leader effectively orchestrate?

Perhaps that will become the true Span of Intelligence.

And the leaders who learn to expand it — without losing the human judgment, accountability, and empathy at its center — may ultimately define not only the future of management, but the future shape of the organization itself.