In a September 17 whitepaper, the American Management Association reported a sharp perception gap in the AI workplace: 62% of managers said their engagement increased over the past year, but 51% of employees said their managers’ engagement was unchanged and 20% said it decreased. For companies pushing AI into everyday work, that matters because the manager’s job may be getting more demanding even as it becomes less visible to the people management is supposed to help.
The survey, based on 1,088 professionals across seven global regions, does not show that AI caused the gap. But it does frame the right business question. Leaders do not need to know only whether managers are using AI. They need to know whether AI is making managers more effective, or simply moving more of their effort into review, judgment and coordination that employees rarely experience directly.
Why employees may miss the work managers say they’re doing
AMA’s findings suggest that managers and employees may be using different definitions of engagement. Employees tend to recognize engagement through visible acts. In the survey, 53% of employees identified open and regular communication as a sign of an engaged manager. Another 42% cited clear direction and expectations, while 37% pointed to listening to employee concerns and supporting employee growth.
Those are not abstract preferences. They are the parts of management employees can actually feel: the manager who explains a change, clarifies priorities, gives feedback, or makes time for a problem before it becomes a failure.
AI can pull managerial effort away from those signals without reducing the amount of work a manager is doing. AMA reported that 36% of managers use AI for routine or administrative work, while the same share said they now spend more time reviewing AI outputs. Another 31% said AI changed the kinds of tasks they delegate, and 30% said they rely more heavily on AI-generated data and insights when making decisions. At the same time, 41% said AI has raised their expectations for how much work employees should complete in a given period.
That is a meaningful shift in the operating model of management. A manager who once spent time drafting, summarizing or handling basic administrative work may now spend more time checking AI-generated material, handling exceptions, weighing risk, resolving ambiguity and explaining why a decision still requires human judgment. That can be real managerial work. It can also be hard for employees to see.
The danger is obvious. If communication and coaching become thinner while review and decision work expands, managers may sincerely feel more engaged even as employees experience less leadership. And more review is not automatically a gain. In some workplaces, it may be valuable oversight; in others, it may amount to rework caused by unreliable systems.
Measure management quality, not just AI activity
This is where many companies risk measuring the wrong thing. Tool access, logins and usage rates may show that AI has been deployed. They do not show whether management improved.
A useful scorecard now needs two layers. The first is visible management behavior: how often managers communicate, whether expectations are clear, how much coaching time they protect, and whether they explain the reasoning behind AI-assisted decisions. AMA explicitly recommends making that cognitive work of management more visible and teaching managers to combine critical thinking with AI rather than treating AI as a substitute for judgment.
The second layer is operating performance: decision quality, output quality, rework, workload, employee trust, team turnover, and how often AI-related errors need escalation. Those are the metrics that can show whether hidden managerial effort is paying off.
That distinction matters most where expectations are rising. If 41% of managers say AI has increased how much work they expect employees to complete, leaders need to test whether staffing, training and review capacity rose with those expectations. Otherwise, a claimed productivity gain can become overload in disguise. Employees may be asked to move faster while managers spend more time validating questionable outputs, resolving mistakes and managing pressure that the organization still counts as efficiency.
Support matters, but visibility still has to be earned
A separate 2026 Gallup workplace study points in a similar direction. Among U.S. employees in organizations that had begun implementing AI, frequent AI use was reported by 79% of employees who strongly agreed their manager supports AI use, versus 46% of those who did not strongly agree. Gallup also found that employees who strongly agreed their manager supported AI were much more likely to say AI transformed work and created opportunities to do what they do best.
That is not proof that manager support caused those outcomes. But it does reinforce the practical point: the manager is still the translator between AI policy and employee experience. Even when AI does not replace managerial accountability, it can blur where that accountability is happening.
AMA’s whitepaper leaves important questions open. The public summary does not spell out the full questionnaire, subgroup breakdowns, weighting method or whether manager and employee responses were matched within the same organizations. It also does not link the engagement figures to retention, productivity, quality, absence or financial performance.
That makes the report less a verdict than a management test. A company can improve perceptions with more check-ins and still make poor decisions. It can also do serious review and judgment work while seeming absent to employees. The real task is to redesign the manager scorecard so both sides of the job count: the visible behaviors employees rely on, and the less visible decision work AI has expanded.
If leaders want to know whether AI is helping managers, they should be able to show where the extra effort lands: clearer expectations, better decisions, stronger coaching, manageable workloads and higher trust. If they cannot, then AI may not be producing better management. It may just be hiding more of it.
By
By
By
By
