Do we need middle managers in the age of AI?

Middle managers are rarely held in high esteem. They’re often referred to as “useless pen pushers,” who don’t really add any value – those above decide on strategy, and those below do the work.

It’s likely that few tears were shed when Android announced significant layoffs across its cloud and Android divisions despite revenue growing apace. It seemed to be a clear signal that AI was coming for middle management.

The reality is not quite as clear-cut, however, and the question isn’t so much whether AI will remove middle management, but whether it will transform what middle management actually means.

That’s not to say there isn’t a link between AI and flatter management structures. For instance, research from UC Berkeley found that there was a clear link between investments in AI and a flattening of the org chart. The researchers argue that you should hire talented individuals proficient in AI, as they need less supervision.

Why pay for oversight when your engineers can manage themselves?

This is notable, as despite accusations that the tech companies are using AI investments as a shield to mask wider layoffs under the guise of innovation, the Berkeley researchers found that AI investments did signal a change in workforce composition.

They found that most firms actually increased headcount after investing in AI, but the mix of those jobs changed considerably. The study showed that college-related jobs rose, as did those in STEM-related roles. Middle management jobs bore the brunt, however.

Not always simple

The problem is that AI creates complexity even as it promises simplification. Routine tasks disappear, yes, but managers must now orchestrate hybrid teams of humans and algorithms. The technology requires constant monitoring.

New forms of routine work emerge – checking AI outputs, calibrating systems, managing edge cases. And employees, anxious about their futures, need reassurance and leadership in ways that no algorithm can provide.

"The productivity of work grows when routine tasks can be passed on to artificial intelligence," researchers from the University of Eastern Finland observed.

"On the other hand, a fast pace of change makes work more demanding."

Managers can focus more on innovation and development, but variation in assignments increases. It is not a reduction in management but a redistribution of what management entails.

More troubling is the social ambiguity. Middle management is fundamentally a relational function, but AI scrambles the usual categories. Is the system a tool or a colleague? Some teams gave their AI systems names, debating who might be the "mother or father" of the technology. These are not merely semantic games, as the uncertainty affects how people interact, collaborate, and understand authority.

Research from Wharton suggests this matters more than organizations realize. In knowledge-intensive industries, such as biotech, computing, and the media, middle managers have an outsize impact on firm performance, often greater than the innovative workers themselves.

Their value lies in project management, resource allocation, and coordination. Strip away middle managers in the name of efficiency, and you risk losing the connective tissue that makes innovation possible.

Avoiding the invisible cage

The experience of gig economy platforms offers a cautionary tale. Consider the freelancing platform examined by researchers at the Kellogg School of Management, where an opaque algorithm determined worker ratings and, by extension, livelihoods.

When the company shifted from transparent metrics to black-box calculations, anxiety spiked. Even top performers could not decipher what made scores rise or fall.

The result was what the researchers termed an "invisible cage," whereby workers were constrained by evaluations they could neither understand nor influence. Some experimented desperately to game the system. Others fled the platform entirely, moving work offline to escape algorithmic judgment. Those most dependent on the platform for income were most trapped.

This is precisely what human middle managers, at their best, prevent. They provide feedback, hear grievances, and explain decisions. When Penn State researchers examined Uber drivers' experiences with algorithmic management, the pattern was clear: drivers appreciated efficiency when things went smoothly but felt powerless when problems arose.

They had no one to appeal to, no means of influencing their working conditions. The platform is optimized for riders, not drivers.

"All of Uber's different management decisions are embodied in the platform," the researchers noted.

The AI did what managers do. It assigned tasks, monitored performance, dispensed discipline, but was stripped of the reciprocal relationship that makes management tolerable. Drivers who joined for autonomy found themselves subject to invisible control.

What comes next

Whether lessons from 2010-2018 apply to the current moment is debatable. Generative AI has accelerated industry concentration, with a handful of firms commanding the data, computing power, and talent necessary to compete. Alphabet's layoffs, despite soaring revenue, suggest that AI-era economics may favor lean operations over employment growth.

Yet the skills required to manage in this environment are expanding, not contracting. Managers need technical fluency to work alongside AI systems. They need emotional intelligence to navigate worker anxiety. They need the judgment to balance algorithmic recommendations against human context.

The Finnish study concluded that AI systems "cannot yet take over all human management in areas such as the motivation and inspiration of team members."

The real issue may be that we are asking the wrong question. The debate should not be whether organizations need middle managers, but what middle managers must become.

The boring, bureaucratic aspects of the role, such as monitoring timesheets, enforcing standard procedures, and shuffling paperwork, are precisely what AI can handle. What remains is harder: ethical judgment when algorithms fail, relationship-building when teams fracture, sense-making when change accelerates.

This requires investment, not elimination. Companies deploying AI without strengthening management risk creating workplaces where efficiency increases, but fairness evaporates. Where productivity rises, but trust collapses.

The invisible cage is not an inevitable feature of algorithmic management. It’s a design failure, and one that human managers, properly supported, can prevent.

The age of AI does not spell the end of middle management. It demands its reinvention.

Unlock more exclusive Cybernews content on YouTube.