The core responsibility of managers is shifting from supervising people to designing an environment in which performance can be produced. As AI and automation take on parts of execution, and as hybrid work and project-based collaboration become commonplace, it is becoming increasingly difficult to move an organization forward through instructions and monitoring alone. In the future, capable managers will not be those who make the most decisions, but those who design goals, authority, information flows, and collaboration structures so that employees can make better decisions themselves.

***

For a long time, the role of managers was clear. They assigned goals set by senior leadership to team members, monitored work progress, and evaluated results. Organizations operated hierarchically, and information flowed according to rank. Managers served as intermediaries who interpreted instructions from above and passed them downward while collecting performance results from the field and reporting them upward. Numerous meetings, reports, and approval procedures were created around this management model.

This model was effective when the nature and path of work were relatively stable. It was clear what had to be produced, who was responsible for each task, and which criteria would be used to assess performance. Managers only had to correct deviations from established procedures, encourage employees who fell short of their goals, and narrow the gap between plans and actual results.

Work today, however, is no longer that simple. Customer demands change rapidly, while multiple departments and external experts participate in a single assignment. A work method that was effective yesterday may become inefficient today. Projects and task forces have become more important than formal organizational structures, while hybrid work across offices and remote locations has become a standard operating model. The very assumption that managers can directly observe and control every work process is beginning to collapse.

The spread of AI has accelerated this change. AI has begun to perform some of the tasks that managers previously reviewed and coordinated, including document preparation, data organization, market research, scheduling, and customer inquiry handling. According to Microsoft’s 2026 survey, employees in organizations with advanced AI adoption were more likely to report that their managers established clear quality standards for AI, created room for experimentation, and actively encouraged work redesign. This indicates that the value of managers lies not in monitoring whether employees use AI, but in designing new ways of working that incorporate AI.

The crisis facing performance managers does not mean that managers are no longer necessary. It means that management itself has become more difficult. In the past, it was enough to assign work to people and check the results. Today, managers must also determine how people and AI should divide responsibilities, what standards different departments should use when collaborating, and how much autonomy should be granted to balance rapid execution with risk control.

If managers continue trying to issue instructions and approve every decision, the organization cannot move faster than the manager’s own processing capacity. As the number of team members increases, reports and approval requests accumulate. Managers become busier, while the team’s ability to make decisions weakens. A paradox emerges in which more intensive management makes the organization increasingly dependent on the manager. What is being shaken is not the managerial position itself, but the outdated model that regards control and monitoring as the essence of management.

When performance is poor, organizations often look first at the people involved. They ask whether the employee lacked competence, had a weak sense of responsibility, or failed to understand the goal clearly. However, when the same problem is repeated by different people, the system must be examined before the individual. Even capable employees struggle to produce strong results in a structure where goals are ambiguous, decision-making authority is unclear, necessary information arrives late, or departmental interests conflict.

A manager acting as a system designer observes how work flows before attempting to control employees. The manager identifies where work begins, who handles it at each stage, and which standards determine its completion. The manager looks for delayed decision points, processes in which the same materials are repeatedly produced, work at organizational boundaries for which no one takes responsibility, and situations where information is concentrated in the hands of a particular individual.

Suppose, for example, that customer complaints are being handled too slowly. A performance manager is likely to examine the number of cases processed by each employee and demand faster responses. A system designer asks different questions. Does the employee have the authority to resolve the customer’s request? Is there a defined response time when cooperation is requested from another department? Can recurring problems be classified automatically? Are previous solutions stored in a searchable format? The system designer looks first for structural factors that delay processing rather than questioning the employee’s attitude.

This approach does not eliminate individual responsibility. It makes responsibility clearer. In an organization that issues vague instructions and evaluates only the results, disputes over who is responsible for failure will repeatedly occur. In a system where roles, authority, decision criteria, and collaboration rules are clear, it becomes evident who should make each decision. Responsibility becomes the result of design rather than the language of pressure.

A good system is not one with a large number of rules. Decisions involving low uncertainty and easily reversible consequences should be made quickly by employees, while only decisions involving significant risks or consequences that are difficult to reverse should be reviewed at a higher level. Requiring managers to approve everything may appear safe, but it can concentrate decisions in the hands of the person with the least direct information. People closest to the work should make judgments, while risks affecting the entire organization should be controlled through shared standards.

The questions managers ask must also change. “Why did you fail to meet the goal?” should become “What structural factor prevented the goal from being achieved?” “Who made the mistake?” should become “What should we change to prevent the same mistake from recurring?” Pressuring people may temporarily increase speed, but improving the system can enhance the speed and quality of future work as well.

Many organizations spend considerable time creating sophisticated goals. They establish figures for sales, profits, customer numbers, production volume, and project completion rates, and distribute them among departments and individuals. However, clear goals do not automatically produce clear execution methods. If employees who receive the same goal operate according to different priorities, or if departments interpret performance criteria differently, goals can intensify conflict rather than promote cooperation.

System designers establish operating principles alongside goals. They determine which tasks should be handled first, who should be notified when a problem occurs, how much can be decided in the field, and what information should be available to every employee. Before demanding performance, they create a path through which performance can be produced.

This requires visibility into work. Visibility does not mean surveillance that allows managers to observe every action taken by team members. It means enabling everyone to see the shared goal, current progress, major obstacles, and the person responsible for each task. When work is visible, less managerial intervention is necessary. Team members can review one another’s progress and determine for themselves when collaboration is needed. When a problem arises, they can request the necessary support without waiting to submit a report.

Meetings must also be treated as part of the performance system. Routine meetings with unclear purposes consume employees’ concentration time while pretending to facilitate information sharing. Information that only needs to be communicated should be shared through documents or collaboration tools. Meetings should be held when conflicting opinions must be reconciled or decisions must be made. When a meeting ends, it should leave behind a record of the decisions, responsible individuals, and deadlines. A meeting that produces no decision is not communication but a delay in work.

The rhythm of feedback is equally important. Waiting until the annual performance review to address problems is too late to change behavior. Yet if managers intervene at every moment, management becomes micromanagement. Managers need a balance in which short, regular check-ins are used to identify gaps between goals and actual conditions without deciding every solution on behalf of employees. Rather than providing all the answers, managers should ensure that the team does not overlook the right questions.

Performance should come from a repeatable process rather than the heroic achievement of a single outstanding individual. If work stops whenever a particular employee goes on vacation, quality fluctuates every time responsibilities are reassigned, and the manager must personally resolve every problem, the organization is not sustained by performance but by individual sacrifice. The performance of system designers should be measured not by how many problems they personally solve, but by how infrequently the same problems recur.

During the early stages of AI adoption, attention focused primarily on improving individual productivity. AI was used as an assistant to write emails faster, summarize reports, and generate ideas. Once AI enters the work process itself, however, the challenge becomes much more complex. Organizations must determine who assigns work to AI, who reviews the output, who is responsible when errors occur, and what data AI is permitted to use.

Managers in the AI era do not need to become technology experts, but they must be able to break work down and reconstruct it. Instead of assigning an entire task solely to either a person or AI, managers should divide it into exploration, drafting, analysis, judgment, approval, and execution, and then assign each stage to the most appropriate party. Repetitive and clearly defined processes can be given to AI and automation, while value judgments, exception handling, relationship building, and final accountability remain with people.

Quality standards matter more than speed in this process. Even if AI produces results quickly, productivity has not truly improved if those results contain factual errors or cause bias, privacy violations, or copyright problems. Managers must determine not only which tasks can use AI, but also what level of quality will be considered acceptable. They should specify what must be fact-checked, which stages require mandatory human review, and what approval procedures must be completed before materials are released externally.

Microsoft’s 2025 survey found that a significant number of leaders expected their teams to train or manage AI agents and build multi-agent systems capable of handling complex tasks within five years. If this forecast becomes reality, managerial teams will include not only people but also multiple digital agents. Managers will no longer assign work only to humans. They will compare the capabilities, costs, and risks of people and AI to create the most appropriate combination.

There is also an expectation that AI will reduce the burden of management, but without adequate design, the opposite can occur. The number of reports and messages may increase, the volume of materials requiring review may surge, and different departments may introduce overlapping automation tools. In such cases, AI creates more work instead of reducing it. The standard for AI adoption should therefore consider not only “What can we do now?” but also “What can we stop doing?”

Managers must measure changes in workload when introducing AI. They should compare processing time, error rates, rework rates, customer satisfaction, and employees’ review burden before and after automation. Using AI is not itself a performance outcome. Performance is achieved when the quality and speed of the overall system improve.

Managers often talk about new strategies while retaining outdated operating methods. They emphasize autonomy while requiring approval for every decision, demand collaboration while evaluating only departmental goals, and call for innovation while penalizing unsuccessful experiments. Employees respond to systems rather than slogans. What an organization genuinely values is revealed through its meeting practices, reporting procedures, and evaluation criteria.

Meetings should be organized around the problem that needs to be solved rather than the rank of the participants. Managers should determine whether everyone truly needs to gather at the same time, what materials should be shared in advance, and what decisions must be reached during the meeting. Instead of dominating the discussion, managers should create conditions in which different information and viewpoints can surface. If the manager’s opinion is presented first, employees are likely to search for answers that match it rather than explore the problem openly.

Reporting must also shift from “informing the manager” to “providing information that supports organizational judgment.” Organizations should reduce processes that require the same content to be rewritten in multiple formats or figures already entered into a system to be copied into another report. Metrics that need to be monitored regularly should be shared through automated dashboards. Reports should focus less on the numbers themselves and more on the causes of change, anticipated risks, and decisions that need to be made.

Evaluation systems should consider both individual performance and contributions to the system. Collaboration cannot function if employees receive high evaluations for achieving their own goals by withholding information or transferring costs to other departments. Conversely, if contributions such as improving work procedures, helping colleagues develop, and reducing recurring problems remain invisible, the organization will continue to be trapped in individual competition.

In the AI era, organizations must distinguish between “how much was produced” and “what value was created.” AI makes it easy to increase the quantity of documents, content, and analytical materials. If the number of outputs remains a performance indicator, organizations will mass-produce low-value results. Evaluation standards must move from output volume to customer impact, improved decision-making, faster problem resolution, and reduced rework.

The way managers themselves are evaluated must also change. Assigning goals to team members and giving them scores is not enough. Managerial responsibility should include whether goals were adjusted to reflect reality, whether necessary resources were provided, and whether obstacles in the collaboration process were removed. A manager who attributes every team failure to individual employees may serve as a performance evaluator, but cannot become a system designer.

A team in which all work stops when the manager is absent may appear well managed. This is because all information and decisions are concentrated in the manager. However, this is not strong leadership but a fragile structure. The moment the manager’s decision is delayed or the manager moves to another position, the speed and quality of the entire organization decline.

Good managers do not reduce their influence. They change the way their influence operates. Instead of issuing direct instructions, they share decision-making criteria. Instead of resolving problems on behalf of employees, they strengthen employees’ ability to solve them. Instead of monopolizing information, they ensure that the right people can access it at the right time. The manager’s knowledge and experience should not remain solely in one person’s mind but should be embedded in the organization’s operating practices.

Three foundations are required. The first is clarity of direction. Employees need to understand why the team exists and which customer problems it should prioritize. The second is clarity of authority. It must be clear who can make which decisions and under what circumstances higher-level approval is required. The third is continuity of learning. Short reviews that document the causes of successes and failures and apply those lessons to future work must be repeated.

Managers must resist the temptation to become the “chief problem solver” who resolves every issue facing the team. When managers provide answers quickly, the arrangement may initially appear efficient, but employees lose opportunities to make judgments themselves. Whenever a similar problem arises, they return to the manager. Instead of asking, “How should I handle this?” managers should ask, “What do we need so that this problem can be resolved next time without depending on any particular individual?”

As this transformation spreads, five changes are likely to emerge in the role of managers.

First, managerial evaluation will move from compiling individual results to measuring system performance. In addition to team members’ goal attainment rates, decision-making speed, rework rates, collaboration delays, employee development, and customer problem resolution times will become key managerial indicators. Managers who credit team members when performance is strong and improve the structure when problems recur will receive higher evaluations.

Second, managing workflows will become more important than managing organizational charts. As the right people are connected quickly for each project and responsibilities change dynamically, the boundaries of fixed departments will weaken. Managers will design networks of work involving multiple departments, external partners, and AI agents rather than controlling only the employees formally assigned to them.

Third, AI agent management will become a formal managerial competency. The ability to handle task instructions, output verification, data access permissions, error responses, and standards for final human approval is likely to be included in leadership development. AI capability will be evaluated as a team operating competency rather than merely an individual’s proficiency with a tool.

Fourth, continuous work redesign will become more prominent than periodic evaluation. The long cycle of setting goals at the beginning of the year and evaluating them at the end will not be sufficient to keep pace with market and technological change. Operating rhythms that adjust goals, roles, and workflows in shorter cycles will spread, and managers will spend more time on design and review than on performance evaluation interviews.

Fifth, the quality of middle managers’ roles will matter more than the number of middle managers. Simple communication and data compilation tasks may be automated or reduced. At the same time, the value of managers who translate strategy into the language of the field, coordinate conflicts between departments, and improve employees’ judgment will increase. Middle managers will not disappear. Administrative managers will decline, while organizational designers will become more important.

The World Economic Forum has forecast that approximately 39 percent of workers’ core skills will change by 2030. Alongside technological capabilities, human skills such as analytical thinking, resilience, leadership, and collaboration are expected to remain important. This means managers cannot remain people who merely evaluate existing capabilities. They must anticipate the capabilities that will be needed in the future and design roles and experiences that allow employees to develop them through real work.

Managerial authority will also come from the reliability of the system rather than rank or control over information. In an organization where goals are clear, decision-making standards are consistent, and operating methods improve through failure, employees do not waste energy trying to guess the manager’s intentions. They know how far they can exercise independent judgment and whom to ask for assistance when support is needed.

Performance managers examine today’s numbers. System designers build the conditions that will produce the next round of performance. Performance managers try to move people, while system designers make work flow properly. The future of management does not lie in tighter control. It lies in connecting people and technology, goals and authority, and autonomy and accountability so that they do not conflict. Work continues even when the manager is absent, and the organization evolves in better ways when the manager is present. That is the new standard of management.

Microsoft WorkLab, May 2026, Microsoft Work Trend Index Research Team, Agents, Human Agency, and the Opportunity for Every Organization

Microsoft WorkLab, April 2025, Microsoft Work Trend Index Research Team, 2025: The Year the Frontier Firm Is Born

World Economic Forum, January 2025, World Economic Forum, The Future of Jobs Report 2025

Gallup, April 2026, Gallup, State of the Global Workplace: 2026 Report

McKinsey & Company, February 2025, Asmus Komm, Fernanda Mayol, Neel Gandhi, Sandra Durth, and Jasmin Kiefer, A New Operating Model for People Management: More Personal, More Tech, More Human