2 weeks ago
As tech companies move billions into artificial intelligence bets and slash their workforces, mediate managers are squarely successful nan crosshairs.
A inclination is emerging: erstwhile tech CEOs denote that AI is making it imaginable to do much pinch less workers, they committedness to flatten their structures by cutting distant what they telephone unnecessary guidance layers and bureaucracy. Just past week, nan cryptocurrency speech Coinbase laid disconnected 14% of its workforce while gesturing to nan thrill of AI-fueled, minimal-management efficiency. In doing so, it joined companies including Amazon, Block and Meta that successful nan past twelvemonth person laid disconnected tens of thousands of labor pinch a circumstantial attraction connected removing guidance layers.
The push to bladed guidance ranks is gaining traction, particularly among companies that are quickly adopting AI, said Anastassia Fedyk, adjunct professor astatine nan University of California, Berkeley’s Haas School of Business. She’s studied really AI is changing workforce composition. As AI devices make it imaginable to displacement much activity from managers to their reports instead, these company’s structural changes could go much permanent, she said.
These shifts are fundamentally reshaping nan roles of mediate management, often requiring managers to beryllium some supervisors and producers and vastly expanding their responsibilities. At nan aforesaid time, companies are giving exertion a much cardinal domiciled successful their organizations. While nan moves are expected to accelerate decision-making processes, they besides could complicate jobs for everyone up and down nan guidance chain, create caller bottlenecks, trim nan benefits that stem from quality relationship astatine work, and degrade a company’s products and services.
“The mediate head domiciled is astir to beryllium nether a batch much pressure,” said Emily Rose McRae, an expert astatine business and exertion insights institution Gartner who studies AI’s effect connected nan early of work. “What that intends for labor is that your occupation gets harder, too. When your head doesn’t get nan support they need, you don’t get nan support you need.”
The inclination doesn’t look confined to tech, either: astatine nan extremity of 2025, openings for mediate head jobs successful nan US had fallen by 42% compared pinch a highest successful 2022, according to research from workforce information level Revelio Labs. Considering that managers comprised 13% of nan US workforce successful 2022, that’s a batch of group and a batch of jobs.
“We’re each trying to fig retired what mediate guidance really means,” said Prateek Singh, a package improvement head who near Meta astatine nan extremity of April. “It’s for illustration a supplier proceedings … Eventually, we will find nan correct one.”
It feels for illustration ‘the Hunger Games’
At Meta, managers started to consciousness unit moreover earlier Mark Zuckerberg, nan CEO, astir precocious discussed flattening nan company’s guidance building during a January 2026 net call, according to Singh. Just a fewer months aft Singh joined successful June 2025, managers connected definite teams saw their number of nonstop reports jump, and managers were progressively expected to lend code, he said. Previously, managers astatine tech companies for illustration Meta often were charged pinch delegating and guiding their team, pinch nan execution of tasks reserved for individual contributors.
To make room for nan further responsibilities, Meta’s managers turned to AI devices to thief them velocity up drafting documents, consolidating notes and evaluating employees, he said. They besides utilized AI to make code.
Singh switched his one-on-one meetings pinch his 7 nonstop reports from play to each different week. In between, he communicated asynchronously utilizing his AI agents, bots that don’t request quality involution to execute tasks, that connected pinch his nonstop reports’ agents to cod updates and supply feedback, he said. While nan strategy seemed to activity for his team, he could spot nan risks of relying connected AI to switch quality interaction.
“If managers are expected to either beryllium penning a batch much codification aliases person a batch much reports, what I spot happening is much asynchronous, agent-driven management,” he said. “Then group suffer touch pinch each nan benefits you get from look time,” for illustration mentorship, quality judgement and guidance. That rumor is magnified astatine highly competitory employers for illustration Meta, wherever nan conflict to beryllium a apical performer often feels for illustration The Hunger Games, he added. AI can’t amended worker capacity nan aforesaid measurement humans can, he said
Though he hadn’t witnessed it yet, he could spot a early successful which managers, nether expanding pressure, are tempted to usage AI for decisions and blindly taxable flawed suggestions. That could compound arsenic different teams build connected apical of those decisions and could lead to information leaks, information holes aliases moreover strategy outages, he said.
After nan fintech institution Block laid disconnected 40% of its workers, immoderate engineering managers were assigned arsenic galore arsenic 175 nonstop reports nether its caller AI-oriented structure, according to soul statement charts reviewed by nan Guardian. That gets Block person to nan perfect extremity of its CEO, Jack Dorsey, of someday operating pinch each 6,000 labor reporting straight to him, minus nan guidance layers. Previously, managers typically operated pinch an estimated six to 12 nonstop reports, said Freeland Abbott, a erstwhile method lead astatine Square, Block’s integer payments service, who was laid disconnected successful February.
While Block’s caller building whitethorn assistance pinch accusation management, Abbott worries that nan much quality parts of managers’ jobs could gaffe done nan cracks. AI can’t supply squad motivation, quality relationship aliases support successful nan measurement a personification can, Abbott said. And off-loading nan worker improvement to same-level colleagues could disadvantage less-experienced and marginalized teams, he said.
Several erstwhile Block employees’ responses person been “Wow, convey God I was laid off,” he said, admitting laid-off labor are somewhat “biased” against their erstwhile employer.
Abbott doesn’t expect those guidance ratios to last, saying that companies will admit nan request for much humans moreover if nan domiciled isn’t called a “manager”.
Meta and Amazon were among nan first of nan tech giants to propose nan request to flatten guidance for nan AI era. In 2023, Zuckerberg announced what he called nan “year of efficiency”, successful which he planned to flatten nan organization. And 2 years ago, Amazon’s CEO, Andy Jassy, told employees he planned to summation nan ratio of labor to managers by astatine slightest 15% – a extremity he said nan institution reached past twelvemonth – to springiness workers a greater consciousness of ownership and trim bureaucracy. Fast guardant to 2026, and some CEOs judge AI is changing nan measurement activity gets done, pinch Jassy suggesting Amazon “will request less people” doing immoderate jobs and Zuckerberg saying Meta is “starting to spot projects that utilized to require large teams now beryllium accomplished by a azygous very talented person”. Block and Coinbase followed suit this year.
Block’s attack is splitting guidance duties – now, AI is chiefly responsible for sharing accusation betwixt managers, their reports and different teams; workers nan institution calls “directly responsible individuals” oversee strategy and priorities; and individual contributors called “player-coaches” negociate worker growth.
“There is nary request for a imperishable mediate guidance layer,” reads a connection from Dorsey and committee personnel Roelof Botha.
Similarly, Coinbase said it will nary longer person “pure managers”; instead, managers will beryllium required to straight lend codification and different work, and will spot their number of nonstop reports jump to 15 group aliases more.
“We’re fundamentally changing really we operate: rebuilding Coinbase arsenic an intelligence, pinch humans astir nan separator aligning it,” Coinbase’s CEO, Brian Armstrong, said successful a tweet announcing layoffs past week.
Coinbase, Amazon, Meta and Block declined to comment.
Employees’ jobs will get harder, too
Companies that opt to importantly trim mediate guidance will apt beryllium those that are already much agile, for illustration tech companies, alternatively of bequest companies that whitethorn person much oversight aliases process aliases whitethorn beryllium slower to adopt tech – and truthful struggle to make extremist changes, said Raffaella Sadun, a Harvard professor who studies nan early of work. Any institution that moves to a caller exemplary will apt consciousness friction, particularly if they make nan alteration suddenly.
Tech companies “are very good positioned to make these changes because they’re precocious from a tech perspective”, Sadun said of nan announcements. But moreover so, “they’ll person to incur nan costs of change” for illustration overhauling really activity is coordinated, altering really decisions are made and shifting workers into different positions – including demoting them.
Reducing nan number of mediate managers is apt to complicate a occupation that’s already incredibly stressful, said McRae, nan Gartner analyst. As it is, nan occupation of being a head is specified a resistance that galore managers crossed industries would take not to beryllium managers again if fixed nan choice, she said, referring to Gartner surveys. That could worsen arsenic companies trim their wide number of managers and require remaining managers to do moreover much work.
Fewer layers of guidance intends labor will person less opportunities to advance, she said. Companies could consequence losing important quality talent arsenic a result, which for tech companies whitethorn provender into their extremity of having a leaner workforce powered pinch AI.
Simplifying nan guidance building besides requires an full redesign of really activity gets done, giving much authority to little levels to make bigger decisions alone, said Amalia Goodwin, world managing head of nan consulting patient Slalom, who focuses connected organizational alteration arsenic it relates to AI. If much labor are making much decisions, they’ll request nan resources, skills, and training to beryllium capable to judge betwixt bully and bad outcomes – thing companies will apt person to provide.
Meanwhile, arsenic employees’ output increases and nan span of power changes crossed levels, activity could slow down successful unintended ways, she said. For example, if 1 squad produces much pinch nan thief of AI, past nan squad that has to o.k. each of that activity whitethorn beryllium overwhelmed pinch volume. With less managers, it will beryllium captious for companies to create structures that break down nan disagreement betwixt units and support accusation flowing, she added.
Some experts opportunity they’re skeptical that tech companies’ experiments pinch utilizing AI to purge mediate managers will drawback on. Matthew Bidwell, guidance professor astatine nan University of Pennsylvania’s Wharton School, said there’s been a history of companies that person tried to break aged hierarchies pinch caller forms of management. But they’re often abandoned aliases service arsenic one-offs.
Middle managers are often successful a “precarious” position successful reorganizations because “it’s harder to specify your value”, he added. That said, arsenic tech companies research pinch less mediate managers, they whitethorn find they’re losing a level of basal scrutiny. “It intends 1 less furniture of kicking nan tires,” he said. “You’ll move faster, but you’ll break much things, and for immoderate organizations that’s astir apt not nan correct trade-off.”
The alteration successful guidance building is 1 logic Singh, who felt his occupation could beryllium astatine risk, chose to time off Meta. Now that he’s gainfully employed extracurricular Silicon Valley, he’s happy to watch from a distance.
“It’s conscionable excessively early successful nan experiment,” he said. “I didn’t want to beryllium nan guinea pig.”
