As tech corporations pour billions into artificial intelligence bets and slash their workforces, center managers are squarely within the crosshairs.
A pattern is rising: when tech CEOs announce that AI is making it attainable to do extra with fewer employees, they promise to flatten their constructions by slicing away what they name pointless administration layers and forms. Simply final week, the cryptocurrency trade Coinbase laid off 14% of its workforce whereas gesturing to the joys of AI-fueled, minimal-management effectivity. In doing so, it joined corporations together with Amazon, Block and Meta that within the final yr have laid off tens of 1000’s of workers with a selected give attention to eradicating administration layers.
The push to skinny administration ranks is gaining traction, particularly amongst corporations which can be quickly adopting AI, mentioned Anastassia Fedyk, assistant professor on the College of California, Berkeley’s Haas College of Enterprise. She’s studied how AI is altering workforce composition. As AI instruments make it attainable to shift extra work from managers to their studies as an alternative, these firm’s structural adjustments might grow to be extra everlasting, she mentioned.
These shifts are essentially reshaping the roles of center administration, usually requiring managers to be each supervisors and producers and vastly increasing their duties. On the similar time, corporations are giving know-how a extra central function of their organizations. Whereas the strikes are anticipated to speed up decision-making processes, in addition they might complicate jobs for everybody up and down the administration chain, create new bottlenecks, cut back the advantages that stem from human interplay at work, and degrade an organization’s services and products.
“The center supervisor function is about to be below much more strain,” mentioned Emily Rose McRae, an analyst at enterprise and know-how insights firm Gartner who research AI’s impression on the way forward for work. “What meaning for workers is that your job will get tougher, too. When your supervisor doesn’t get the help they want, you don’t get the help you want.”
The pattern doesn’t appear confined to tech, both: on the finish of 2025, openings for center supervisor jobs within the US had fallen by 42% in contrast with a peak in 2022, in line with research from workforce data platform Revelio Labs. Contemplating that managers comprised 13% of the US workforce in 2022, that’s lots of people and plenty of jobs.
“We’re all attempting to determine what center administration actually means,” mentioned Prateek Singh, a software program improvement supervisor who left Meta on the finish of April. “It’s like a drug trial … Ultimately, we’ll discover the suitable one.”
It appears like ‘the Starvation Video games’
At Meta, managers began to really feel strain even earlier than Mark Zuckerberg, the CEO, most just lately discussed flattening the corporate’s administration construction throughout a January 2026 earnings name, in line with Singh. Just some months after Singh joined in June 2025, managers on sure groups noticed their variety of direct studies bounce, and managers have been more and more anticipated to contribute code, he mentioned. Beforehand, managers at tech corporations like Meta usually have been charged with delegating and guiding their crew, with the execution of duties reserved for particular person contributors.
To make room for the extra duties, Meta’s managers turned to AI instruments to assist them velocity up drafting paperwork, consolidating notes and evaluating workers, he mentioned. In addition they used AI to generate code.
Singh switched his one-on-one conferences together with his seven direct studies from weekly to each different week. In between, he communicated asynchronously utilizing his AI brokers, bots that don’t want human intervention to execute duties, that related together with his direct studies’ brokers to gather updates and supply suggestions, he mentioned. Whereas the technique appeared to work for his crew, he might see the dangers of counting on AI to exchange human interplay.
“If managers are anticipated to both be writing much more code or have much more studies, what I see occurring is extra asynchronous, agent-driven administration,” he mentioned. “Then folks lose contact with all the advantages you get from face time,” like mentorship, human judgment and steerage. That concern is magnified at extremely aggressive employers like Meta, the place the battle to be a high performer usually appears like The Starvation Video games, he added. AI can’t enhance worker efficiency the identical manner people can, he mentioned
Although he hadn’t witnessed it but, he might see a future wherein managers, below rising strain, are tempted to make use of AI for selections and blindly submit flawed options. That might compound as different groups construct on high of these selections and will result in knowledge leaks, safety holes and even system outages, he mentioned.
After the fintech company Block laid off 40% of its workers, some engineering managers have been assigned as many as 175 direct studies below its new AI-oriented construction, in line with inside group charts reviewed by the Guardian. That will get Block nearer to the best objective of its CEO, Jack Dorsey, of sometime working with all 6,000 employees reporting directly to him, minus the administration layers. Beforehand, managers sometimes operated with an estimated six to 12 direct studies, mentioned Freeland Abbott, a former technical lead at Sq., Block’s digital funds service, who was laid off in February.
Whereas Block’s new construction could support with info administration, Abbott worries that the extra human elements of managers’ jobs might slip via the cracks. AI can’t present crew motivation, human connection or help in the way in which an individual can, Abbott mentioned. And off-loading the worker improvement to same-level colleagues might drawback less-experienced and marginalized groups, he mentioned.
A number of former Block workers’ responses have been “Wow, thank God I used to be laid off,” he mentioned, admitting laid-off workers are barely “biased” in opposition to their former employer.
Abbott doesn’t expect those management ratios to last, saying that corporations will acknowledge the necessity for extra people even when the function isn’t referred to as a “supervisor”.
Meta and Amazon have been among the many first of the tech giants to counsel the necessity to flatten administration for the AI period. In 2023, Zuckerberg introduced what he referred to as the “year of efficiency”, wherein he deliberate to flatten the group. And two years in the past, Amazon’s CEO, Andy Jassy, told employees he deliberate to extend the ratio of workers to managers by a minimum of 15% – a objective he mentioned the corporate reached final yr – to present employees a larger sense of possession and cut back forms. Quick ahead to 2026, and each CEOs imagine AI is altering the way in which work will get performed, with Jassy suggesting Amazon “will need fewer people” performing some jobs and Zuckerberg saying Meta is “beginning to see tasks that used to require massive groups now be achieved by a single very proficient particular person”. Block and Coinbase adopted swimsuit this yr.
Block’s strategy is splitting administration duties – now, AI is primarily liable for sharing info between managers, their studies and different groups; employees the corporate calls “straight accountable people” oversee technique and priorities; and particular person contributors referred to as “player-coaches” handle worker progress.
“There isn’t any want for a everlasting center administration layer,” reads a statement from Dorsey and board member Roelof Botha.
Equally, Coinbase mentioned it can not have “pure managers”; as an alternative, managers shall be required to straight contribute code and different work, and can see their variety of direct studies bounce to fifteen folks or extra.
“We’re essentially altering how we function: rebuilding Coinbase as an intelligence, with people across the edge aligning it,” Coinbase’s CEO, Brian Armstrong, mentioned in a tweet saying layoffs final week.
Coinbase, Amazon, Meta and Block declined to remark.
Staff’ jobs will get tougher, too
Firms that decide to considerably cut back center administration will probably be these which can be already extra agile, like tech corporations, as an alternative of legacy corporations which will have extra oversight or course of or could also be slower to undertake tech – and due to this fact wrestle to make radical adjustments, mentioned Raffaella Sadun, a Harvard professor who research the way forward for work. Any firm that strikes to a brand new mannequin will probably really feel friction, particularly in the event that they make the change instantly.
Tech corporations “are very effectively positioned to make these adjustments as a result of they’re superior from a tech perspective”, Sadun mentioned of the bulletins. Besides, “they’ll must incur the price of change” like overhauling how work is coordinated, altering how selections are made and shifting employees into completely different positions – together with demoting them.
Decreasing the variety of center managers is more likely to complicate a job that’s already extremely irritating, mentioned McRae, the Gartner analyst. As it’s, the job of being a supervisor is such a drag that many managers throughout industries would select to not be managers once more if given the selection, she mentioned, referring to Gartner surveys. That might worsen as corporations cut back their general variety of managers and require remaining managers to do much more work.
Fewer layers of administration means workers could have fewer alternatives to advance, she mentioned. Firms might danger dropping vital human expertise in consequence, which for tech corporations could feed into their objective of getting a leaner workforce powered with AI.
Simplifying the administration construction additionally requires a complete redesign of how work will get performed, giving extra authority to decrease ranges to extend selections alone, mentioned Amalia Goodwin, international managing director of the consulting agency Slalom, who focuses on organizational change because it pertains to AI. If extra workers are making extra selections, they’ll want the assets, abilities, and coaching to have the ability to decide between good and dangerous outcomes – one thing corporations will probably have to offer.
In the meantime, as workers’ output will increase and the span of management adjustments throughout ranges, work might decelerate in unintended methods, she mentioned. For instance, if one crew produces extra with the assistance of AI, then the crew that has to approve all of that work could also be overwhelmed with quantity. With fewer managers, will probably be important for corporations to create constructions that break down the divide between models and preserve info flowing, she added.
Some consultants say they’re skeptical that tech corporations’ experiments with utilizing AI to purge center managers will catch on. Matthew Bidwell, administration professor on the College of Pennsylvania’s Wharton College, mentioned there’s been a historical past of corporations which have tried to interrupt outdated hierarchies with new types of administration. However they’re usually deserted or function one-offs.
Center managers are sometimes in a “precarious” place in reorganizations as a result of “it’s tougher to outline your worth”, he added. That mentioned, as tech corporations experiment with fewer center managers, they might discover they’re dropping a stage of mandatory scrutiny. “It means one fewer layer of kicking the tires,” he mentioned. “You’ll transfer quicker, however you’ll break extra issues, and for some organizations that’s most likely not the suitable trade-off.”
The change in administration construction is one motive Singh, who felt his job could possibly be in danger, selected to depart Meta. Now that he’s gainfully employed exterior Silicon Valley, he’s comfortable to observe from a distance.
“It’s simply too early within the experiment,” he mentioned. “I didn’t wish to be the guinea pig.”