AI Will Change the Economics of Management

A management meeting can spend most of its time establishing what happened. Teams reconcile numbers, explain delays, identify dependencies, and prepare decisions that move to another meeting. Each activity may be necessary. Together, they reveal how much organizational effort goes into making work visible and getting it coordinated.

AI can help assemble that picture more quickly. But a faster report is only the first-order opportunity. If information gathering, analysis, and routine coordination become less expensive, the organization should reconsider the structures built partly to perform them.

That makes management itself a subject of AI transformation. The relevant questions concern how many people a manager can effectively support, what each layer contributes, where decisions belong, and how attention is allocated. They also concern what happens to leadership when routine administrative demands recede.

Normatic's view is that lower coordination costs create an opportunity to redesign management. Whether that opportunity supports wider spans, fewer layers, stronger specialist roles, or more investment in coaching depends on the work and the evidence. Time saved on reporting cannot, by itself, justify a new organization chart.

Hierarchies reflect the cost of coordinating work

Management structures help an enterprise act despite incomplete information, fragmented knowledge, and limited attention. Managers interpret strategy, reconcile competing priorities, distribute resources, and intervene when the work cannot proceed within existing arrangements. Hierarchies also provide routes for escalating problems to people with greater expertise or authority.

Some of this structure reflects the expense of moving information. A supervisor consolidates team updates. A department head translates them into a common view. A business leader decides which issues merit intervention. At each level, reporting and interpretation can be intertwined with substantive decisions.

Established organizational research provides a useful starting point. Bloom, Garicano, Sadun, and Van Reenen found that information technologies were associated with greater autonomy and wider spans in manufacturing firms, while communication technologies were associated with less autonomy. Their explanation distinguishes easier access to knowledge from easier access to a superior's decisions. This is foundational research published in 2014, not evidence about today's AI deployments. It establishes why different information costs can support different organizational designs. Research: information, communication, and firm organization.

AI brings several possibilities into the same system: retrieving relevant knowledge, synthesizing operational information, preparing options, and routing exceptions. Our interpretation is that the economic justification for some management activities may change as those capabilities become reliable within actual workflows. The effect on structure remains a separate question.

The costs also move. Organizations must maintain the underlying information, evaluate outputs, investigate false alerts, and support the systems. A reduction in managerial reporting effort may require additional work in shared services or technical operations. The relevant comparison is the cost and quality of coordination across the enterprise, including errors and delays, rather than the minutes saved by one manager.

Coordination work and leadership respond differently

Management is a bundle of activities. Preparing a performance summary and deciding how to respond to a struggling team are related, but they make different demands on a manager.

AI may reduce the effort needed to collect evidence, identify discrepancies, compare scenarios, or draft a decision brief. Human judgment remains essential when the evidence is incomplete, objectives conflict, or consequences must be owned. Coaching, motivation, conflict resolution, and the relationships through which people cooperate cannot be inferred from a reporting dashboard.

Exhibit 1. Lower preparation costs leave consequential decisions to be made

Illustrative activity analysis. These are possible changes in work, not measured savings or a classification of roles for removal. Source: Normatic analysis.

Emerging AI research supports a bounded claim. In a field experiment involving 776 professionals at Procter & Gamble, individuals using AI matched teams without AI on the studied product-innovation task, and AI-assisted proposals combined technical and commercial perspectives more effectively across professional backgrounds. The study shows how some benefits of collaboration can become more accessible. It does not establish that management roles or coordination across an operating enterprise can be eliminated. Research: The Cybernetic Teammate, published in Organization Science in 2026.

A separate field experiment illustrates the limits of simply providing tools. The November 2025 version of Shifting Work Patterns with Generative AI covers 7,137 knowledge workers across 66 firms. It reports email time savings among treated workers who used the tool, but no detected broader shifts in the quantity or composition of tasks from individual-level provision. Individual efficiency and organizational redesign are different outcomes. Research: Shifting Work Patterns with Generative AI.

Neither study identifies an optimal managerial span or hierarchy depth. The design implications that follow are hypotheses for organizations to test against their own operating conditions.

A wider span must still allow managers to support people

If much of a manager's time goes into obtaining status, checking routine progress, and preparing recurring reports, reliable support could create capacity to lead a larger team. Better visibility may also help the manager intervene earlier, before small issues require prolonged attention.

The constraint may instead be coaching, ambiguity, or relationship intensity. An inexperienced team learning unfamiliar work may need frequent discussion even when its status is perfectly visible. A team negotiating across several functions may require substantial managerial intervention because priorities conflict, not because information is missing.

Consider an illustrative operations team whose manager currently spends substantial time chasing updates and reconciling queues. Automated summaries could release capacity. An organization might use that capacity to increase the span, improve coaching, or resolve recurring problems. These are competing uses of the same managerial attention, and the first is not automatically the most valuable.

Nor are all saved minutes interchangeable. Short intervals released throughout the day may not provide the sustained availability needed for difficult conversations. Additional alerts can consume the apparent saving. Before widening a span, leaders should examine response times, access to coaching, unresolved issues, and outcomes under normal demand and during disruption.

The design objective is effective managerial capacity. Direct reports per manager are only one indicator of it.

A layer should contribute more than information transmission

Some intermediate layers compile reports, relay decisions, and coordinate approvals whose purpose has become difficult to explain. If AI makes those activities faster and more reliable, preserving the layer solely to perform them becomes harder to justify economically.

Other layers hold domain expertise, resolve conflicts across units, manage material risks, or make resource decisions at a scale no individual team can address. Their value may remain substantial even when their reporting work changes. Organizational size and the consequences of decisions still matter.

The starting point should be an examination of what a layer actually contributes. Which decisions does it make? Which problems does it resolve? What expertise or independent challenge does it provide? If it disappeared, who would inherit those responsibilities, and with what capacity?

Removing a reporting layer while sending every unresolved issue to an executive can increase delay and weaken oversight. It can also push coordination work into informal networks where responsibility is harder to see. A shorter chart is not proof of a lower coordination cost.

One possible response is to separate routine consolidation from substantive management. A shared capability could maintain operational information and prepare common analyses, while managers retain responsibility for interpretation and action. Specialist coordination may remain necessary where dependencies span functions. This changes the composition of management work without prescribing a uniform reduction in management roles.

Better information does not determine where authority belongs

AI can give frontline teams enough context to resolve more cases within delegated limits. The same capabilities can give central leaders a clearer view of operations and make it easier to intervene. Greater delegation and greater central visibility can coexist; central visibility need not mean central approval.

Consider an illustrative network of service teams handling customer exceptions. One design gives local managers the evidence and authority to resolve qualifying cases, with central monitoring for patterns. Another sends supported recommendations to a central decision owner because the cases create commitments across the enterprise. Both can use the same information infrastructure.

Exhibit 2. Shared information supports different allocations of authority

The choice depends on where relevant knowledge resides, how tightly decisions affect other teams, and who can responsibly own their consequences. A local manager may understand a customer relationship better. A central owner may be better placed to allocate a scarce resource shared across regions.

Cheaper access to information does not remove the time needed to make a decision. If every new signal generates a central approval request, improved visibility can create an overloaded decision queue. Organizations need explicit thresholds for escalation and intervention, alongside clear boundaries for local action.

Decision speed therefore depends on both information and authority. A well-prepared recommendation still waits if nobody has the mandate or capacity to act.

Management attention may concentrate on harder work

As routine coordination becomes more reliable, the remaining managerial workload may become more demanding. Ambiguous cases, cross-functional conflicts, resource trade-offs, coaching, and strategic choices can occupy a larger share of attention.

In The Future Organization Runs on Exceptions, we examined how routine flow can leave people handling the cases that require intervention. The management implication is a change in the mix of work. Fewer routine escalations do not necessarily produce a proportionate reduction in the capacity needed to handle the remainder.

Leaders should anticipate that shift in role design. Managers may need stronger skills in interpreting uncertain evidence, challenging system recommendations, negotiating across functions, and developing the judgment of others. They also need time to identify recurring causes and change the underlying process, rather than repeatedly absorb the same exceptions.

The management system can support this work through risk alerts, scenario analysis, resource forecasts, and portfolio monitoring. Its purpose should be to direct attention to decisions that matter. An expanding volume of automatically generated analysis can otherwise consume the attention it was meant to release.

More measurement can produce worse management

AI can make organizational activity easier to observe and summarize. That does not make every observable activity a valid measure of contribution.

A team rewarded for closing cases may avoid difficult ones. A manager receiving a ranked list of employees may mistake differences in assignments or recording practices for differences in performance. A precise score can conceal incomplete evidence, inconsistent definitions, or a model's unsupported inference.

The OECD's 2025 survey of more than 6,000 firms across six countries found that managers perceived benefits from algorithmic management alongside concerns about unclear accountability, difficulty following system logic, and inadequate protection of worker health. These are reported perceptions concerning algorithmic management broadly, not causal evidence that generative AI improves managerial decisions. Research: Algorithmic management in the workplace.

Our view is that management instruments require deliberate limits. The organization should establish what information is appropriate to collect, who may use it, and which decisions it can support. Monitoring should be proportionate to its operational purpose. People need ways to correct inaccurate records and challenge consequential interpretations.

Performance summaries should preserve relevant context and uncertainty. Any use in decisions affecting people requires scrutiny for bias and an accountable human decision-maker with the ability to examine and reject the recommendation. Automated scoring is not a substitute for coaching or fair assessment.

Stronger visibility should also leave room for work that is difficult to count: developing colleagues, preventing problems, building relationships, and exercising restraint. Management loses value when the organization optimizes the available metric instead of the intended outcome.

Redesign should follow a change in operating evidence

The practical starting point is a recurring management process, such as an operational review, resource-allocation cycle, or escalation route. Establish where effort, delay, and errors arise. Introduce AI support for a defined part of the process, then examine whether the constraint has moved.

Exhibit 3. Structural changes need evidence beyond task-level time savings

Decision criteria for a bounded redesign pilot, not a maturity score or a staffing formula. Evaluate outcomes under representative operating conditions. Source: Normatic analysis.

A useful evaluation follows work through to the decision and its consequences. It considers preparation time, decision latency, quality, coordination effort elsewhere, and the support available to people. It also includes the cost of maintaining the AI capability. A locally faster process can still produce a worse enterprise outcome.

Where evidence supports a structural change, responsibilities, incentives, and escalation paths should change with it. Where it does not, the organization may still benefit from giving managers more capacity to lead their existing teams. Both can be valid outcomes.

The competitive opportunity extends beyond distributing copilots. Organizations can use lower information and coordination costs to reconsider how management creates value: where it supplies expertise, where it allocates authority, where it develops people, and where it resolves competing demands.

AI will change some of the work managers perform. Its deeper significance may be that it changes the economics that determine why management is structured the way it is.

Sources

Established organizational research

• Nicholas Bloom, Luis Garicano, Raffaella Sadun, and John Van Reenen, The Distinct Effects of Information Technology and Communication Technology on Firm Organization, Management Science, 2014. Foundational evidence on information costs, autonomy, and spans; used to frame the economic mechanism, not to estimate current AI effects.

Emerging AI and management evidence

• Fabrizio Dell'Acqua and colleagues, The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork, Organization Science, published online June 2026; accessible working-paper record. A bounded product-innovation experiment, not a longitudinal study of management structures.
• Eleanor Wiske Dillon, Sonia Jaffe, Nicole Immorlica, and Christopher T. Stanton, Shifting Work Patterns with Generative AI, working paper, version revised November 13, 2025. The revised sample and findings are used here rather than the earlier release. Individual tool access does not test a deliberate redesign of management.
• Anna Milanez, Annikka Lemmens, and Carla Ruggiu, Algorithmic management in the workplace: New evidence from an OECD employer survey, OECD Artificial Intelligence Papers No. 31, February 2025. Employer survey evidence about perceived effects and governance concerns; algorithmic management includes software beyond generative AI.

Normatic interpretation

The organizational implications, illustrative examples, and three exhibits are Normatic analysis. They identify choices to test and evidence to seek; they do not claim a measured reduction in management headcount or prescribe a universal span, hierarchy depth, or allocation of authority.