If meetings and reports shrink, do managers shrink with them?

Break down a middle manager’s day and a surprising amount of it now overlaps with what AI can do well: collecting updates, turning them into reports, summarizing meetings, assigning next actions and keeping schedules and numbers current. When information was scattered across people and systems, simply gathering it and carrying it upward was a meaningful management function.

That premise is changing. AI can read the digital trail left in workplace tools and generate routine reporting on demand. The obvious question follows: if the organization no longer needs a human information relay, does it still need as many layers of middle management?

Companies are flattening management — but “AI fired the managers” is too simple

In 2026, Block described a major restructuring that would reduce company-wide headcount by more than 40%. In explaining the change, the company said it had spent roughly two years modernizing how it operated, flattening management layers and accelerating product development. That makes organizational thinning a real management decision, not a hypothetical future-of-work scenario.

But the headline does not mean Block cut 40% of middle managers because of AI. The figure covers the company as a whole, and business structure, cost discipline and growth strategy all matter. LinkedIn’s 2026 labor-market work similarly cautions against explaining hiring weakness in advanced economies with AI alone; macroeconomic uncertainty and monetary conditions remain important parts of the picture.

The first management work AI attacks is the relay function

Status consolidation, recurring reports, meeting summaries, schedule coordination and repetitive document preparation are especially exposed because the underlying information is already digital and the rules are relatively explicit. If a management layer mainly exists to move that information around, the case for preserving it at the same scale gets weaker.

AWS has made a similar organizational-design argument: companies adopting AI should revisit unnecessary management layers and decide, level by level, what humans should own and what AI can handle. The manager’s role shifts away from traditional supervision toward mentoring and quality assurance.

Yet advanced AI organizations show a larger management effect, not a smaller one

A separate Microsoft survey of 1,800 workers connected to the 2026 Work Trend Index found that when managers visibly modeled AI use, employees were 17 percentage points more likely to report seeing AI’s value, 22 points more likely to critically evaluate AI output, and 30 points more likely to trust agentic AI.

That does not prove that managers mechanically caused every difference. It does show that once AI enters the organization, someone still has to define where it belongs, what counts as acceptable output and where human review becomes mandatory. There is a difference between distributing an AI tool and designing a working system around it.

Korea shows the same divide in management quality

Microsoft’s 2026 Korea data shows a similar pattern. Among Frontier Professionals, 74% said managers openly use AI, compared with 53% of general respondents. Sixty-nine percent said managers establish quality standards for AI-enabled work, versus 43% in the broader group. The gap was also large for creating an environment where employees could experiment: 72% versus 47%.

These are not causal estimates of what a “good manager” adds. But they do show that in organizations where AI use is more advanced, the manager has not disappeared. The role becomes concrete in visible AI use, quality thresholds, experimentation and work redesign.

Management value shifts from knowing everything to defining the standard

Managers once gained a great deal of leverage simply by being the person who knew what everyone was doing. As AI becomes able to search work logs, summarize status and draft materials, hoarding or routing information becomes a weaker source of differentiation.

What remains is harder: choosing priorities, deciding which output is good enough, defining when an exception must return to a human, resolving trade-offs and owning the final result. AI can threaten a manager whose advantage was “I know what everyone is doing.” It can amplify a manager whose advantage is “I can decide what a good result looks like.”

What should an AI-era manager put on a résumé?

“Managed a team of eight,” “ran weekly meetings” and “reported to executives” describe responsibility, but not management quality. As routine coordination gets cheaper, those lines reveal less about what the manager actually changed.

A stronger record would show how many approval steps were removed, how decision time changed, how AI quality standards were designed, which workflows were automated, what conditions triggered escalation back to a human, and what happened to productivity, customer or quality metrics. These are closer to evidence of management leverage in an AI-enabled organization.

There may be fewer managers. Management is not going away.

As AI and organizational flattening progress, individual managers may cover broader spans and some management positions may disappear. But the more human work is amplified by software and agents, the more organizations must answer who decides what to delegate, what must be reviewed and who owns the final outcome.

For middle managers, the practical move is not simply to defend the title. It is to separate the part of the job that moves information from the part that improves decision quality. The cheaper AI makes the former, the more valuable it becomes to prove the latter.

BANSEOG VIEW

The durable manager is an owner of decision quality, not an information traffic controller

The number of managers may fall in some organizations. That is different from management disappearing. As AI makes information aggregation and routine reporting cheaper, more of the human manager’s value concentrates in standards, exceptions, coaching and accountability.

The market value of a manager may increasingly depend less on span of control and more on evidence of how work was redesigned between people and AI, what quality bar was set, and how the result improved.

SOURCES

Primary sources

The sources measure different things: surveys, company restructuring and labor-market conditions. This article does not combine them into a single causal claim. It uses them to separate management functions that are becoming easier to automate from those that retain or gain value.