Code-named Undertaking OT – quick for Group Transformation – the plan envisioned an “AI native” future for the proprietor of Fb and Instagram. AI would take over a lot of the day by day work carried out by 1000’s of human workers. Digital staff can be overseen inside Meta by smaller, “talent-dense” cadres of human staffers, in response to one inner planning doc reviewed by Reuters and three individuals conversant in the challenge.
In scenario-planning workout routines, two of those individuals stated, executives explored slashing the scale of many groups throughout Meta by as a lot as 60%. Some workers can be supplied roles in new items, whereas others can be laid off as a part of a culling that one human-resources govt projected can be as large as or greater than the corporate’s cuts of round 25% three years in the past, in response to one other inner doc.
The restructuring can be carried out in two “waves,” starting with a primary purge in Could and adopted by one other shake-up in November, inner planning paperwork seen by Reuters confirmed. Layoffs can be supplemented with the closing of open positions and pushing individuals out who Meta believed are poor performers. These and different particulars of the plan, together with the dimensions of the restructuring, have not been beforehand reported. However on the evening of Could 19, simply hours earlier than the primary layoff wave, Zuckerberg blinked. Meta laid off 10% of its workers the following day, however it known as off planning for the November cuts, in response to one inner doc reviewed by Reuters. By then, Meta workers had been in open revolt, satisfied that the corporate’s AI transformation initiatives had been partly geared toward changing them. Inner information was additionally suggesting that autonomous AI “agent” expertise on the coronary heart of the technique was failing to ship hoped-for productiveness good points. Some traders had been questioning what Meta needed to present for its gargantuan spending on AI.
This text reveals for the primary time the fast tempo of the cuts Meta was contemplating, the considering behind the plans and the way they shortly unraveled. Based mostly on scores of inner paperwork, posts and recordings reviewed by Reuters, in addition to conversations with greater than 20 individuals with information of Meta’s internal workings, the reporting reveals how the social-media large tried to place itself on the forefront of an AI-driven office overhaul, solely to stumble within the execution.
In response to this story, Meta confirmed the existence of Undertaking OT, which it described as a year-long challenge targeted on price slicing, redesigning staff buildings and shifting workers into new precedence areas, such as producing coaching information for its AI fashions. The corporate acknowledged the plan was to be carried out in two waves and that probably the most drastic eventualities concerned decreasing the scale of a few of Meta’s groups by as much as 60%. But it surely declined to specify which items had been concerned and stated a number of main ones weren’t a part of it. Meta stated the eventualities included each layoffs and redeployments, and that the corporate at no level supposed to put off 60% of its total workforce. Firm leaders canceled plans for the second wave earlier than figuring out how many individuals general would lose their jobs, it stated. “As a part of our firm restructuring earlier this 12 months, we requested some groups to conduct a situation planning train wanting on the potential impression of redeployments, open function closures and cuts,” Meta stated in an announcement. “This in the end resulted in transferring 1000’s of workers to do precedence work on a number of newly-established groups, as has been publicly reported. In the end, we did not transfer ahead with each situation from the train – and it was by no means assumed we might.”
Reuters was unable to find out what precisely prompted Zuckerberg to shift course or what the social-media large’s present plans are for persevering with to reshape its workforce. Meta declined to make Zuckerberg obtainable for remark.
THE ‘AI-NATIVE PLAYBOOK’
Ever for the reason that launch of OpenAI’s ChatGPT in late 2022, Silicon Valley leaders have been imagining the revolutionary prospects that generative AI may unleash for the way forward for work. They’ve been particularly intrigued by the promise of “agentic” capabilities. Brokers, a nascent expertise, are being designed to do extra than simply spit out solutions to queries like typical AI chatbots. As an alternative, they’re purported to take autonomous actions similar to buying, reserving journey and creating new apps.
At Meta, executives final 12 months threw themselves into AI-inspired administration philosophies. They had been captivated by concepts rising out of the startup world about how firms should not simply incorporate AI into their present processes, however as a substitute go totally “AI native.” As one Undertaking OT doc envisioned it: “AI-ready instruments and brokers work together, workflows are automated, new builds are AI-first.” The technique additionally included having Meta promote AI brokers, which might carry out duties like scheduling appointments and shutting gross sales, to different firms.
Meta executives, together with Chief Knowledge Officer Alex Schultz and Head of Product Naomi Gleit, visited Asia final 12 months and admired how startups there had constructed their organizational charts round AI, in response to three individuals conversant in the journeys. Meta executives additionally commissioned their very own analysis into how AI startups had been organized and arrange pilot initiatives to find out what being “AI native” would imply on the firm, in response to one supply conversant in the analysis and inner paperwork describing the pilots.
Requested in June concerning the genesis of Meta’s AI workforce transformation, Gleit informed Reuters she had spent “fairly a little bit of time” in Meta’s Singapore workplace final 12 months and stated the practices there had “impressed a few of the groups in California and New York.” She stated most of the concepts had been “bottom-up,” reflecting how some Meta workers had been already beginning to implement adjustments themselves.
Ime Archibong, a long-time vp of product administration, introduced one among Meta’s earliest pilot initiatives in July final 12 months. In an inner publish, he described how his staff was taking its first step to modernize product growth, describing the transition in sports activities phrases. “In basketball, embracing the quick break permits you to take extra pictures – and higher ones. We anticipate the identical with AI instruments: they permit us to discover extra concepts with much less price and better constancy,” Archibong wrote. He predicted that prototypes created shortly with AI would extra intently resemble completed merchandise than earlier than. “This is not simply extra enjoyable – in an AI-first period, we imagine it is the profitable technique.”
Zuckerberg, too, alluded to the potential for AI to make work “much more enjoyable” in an earnings name six months later.
In sensible phrases, Archibong’s pilot concerned organising 5 “small tech pods,” every consisting of two to 3 engineers and a designer, all outfitted with AI instruments. The teams would dispense with established processes for releasing new merchandise, similar to fastened six-month planning cycles. As an alternative, they’d goal to construct prototypes in four-week “sprints,” he wrote in his publish.
In October, a member of Archibong’s staff previewed a broader roll-out of the strategy. In an inner publish titled “AI-Native Playbook,” the individual laid out a information for different groups to “make the leap.” Conventional roles for product designers and engineers would vanish, and tech pod members would get a brand new generic title: “builder.” Layers of center administration can be eradicated. As an alternative, pods would report as much as a single, high-level unit head. “Agent-assisted evaluation” would assist set day-to-day priorities. Meta declined to make Archibong obtainable for remark or to reply questions on its early “AI native” pilots.
Early this 12 months, Zuckerberg set Undertaking OT in movement and directed executives to proceed with the adjustments to administration buildings. By June, not less than 11 items, together with engineering and analysis groups, had applied small pods, in response to inner bulletins and 5 individuals conversant in the preparations.
With the adjustments, Meta launched a “village strategy” to managing individuals, in response to an inner announcement from one unit. Job efficiency rankings and promotions can be determined by high-level unit heads, or “Org Leads,” supported by human-resources personnel and unspecified “AI techniques,” the publish stated. Every unit head would oversee between 30 and 50 individuals, whereas “Pod Leads” would steer small pods day-to-day, however don’t have any formal administration authority, the sources and bulletins stated.
One Meta staffer assigned to handle a pod expressed confusion concerning the new setup on an inner firm message board, in response to a replica of the publish. “I am not going via supervisor coaching, and I am not gaining access to rankings & supervisor instruments,” the individual wrote.
Meta informed Reuters that groups “experimented in numerous methods with tips on how to be extra agile.” The corporate declined to clarify the publish’s reference to “AI techniques,” however stated, “Efficiency ranking and promotion selections had been and are made by individuals, not AI.”
Across the similar time, Meta additionally launched a brand new human assets instrument to determine “Irreplaceable Expertise,” in response to an inner doc seen by Reuters. The doc referred to a hypothetical “10X Performer” – a nod to the decades-old Silicon Valley archetype of an engineer who can do the work of 10 individuals. The time period is again in vogue now, as many AI devotees imagine the expertise will allow extraordinarily gifted people to carry out the work of total groups.
In line with two individuals conversant in the matter, Meta supposed to spend a few of its financial savings from layoffs on large pay packages to draw and retain such expertise, particularly AI engineering rock stars.
EMPLOYEE REBELLION
On March 13, earlier than many leaders on the vice-president degree had even been briefed on Undertaking OT, Reuters reported that Meta was planning layoffs that might have an effect on 20% or extra of its workforce. It would not have been the primary culling on that scale: Between late 2022 and early 2023, Meta slashed round 25% of its workers. Nonetheless, the report induced alarm. Many rank-and-file workers had been rattled. They usually weren’t but purported to know concerning the cuts, annoying Zuckerberg and leaving executives unprepared for the backlash, in response to an individual conversant in the matter. A Meta spokesperson on the time known as the story “speculative reporting about theoretical approaches.”
Internally, high-level executives hunkered down. They opted to not talk about the article with the rank-and-file and quietly dismissed it to senior managers, instructing them to inform their groups they need to anticipate roles to “evolve” because of AI, in response to a talking-points doc seen by Reuters.
In April, Reuters printed extra particulars: Meta was set to chop about 10% of its workforce in a primary wave of layoffs on Could 20 and was aiming to shed extra workers within the second half of the 12 months. Meta quickly confirmed the ten% discount plan to workers, and Zuckerberg later informed workers the cuts had been because of heavy capital expenditures.
Executives had began transferring forward with different elements of the plan. Some engineers, for instance, had been reassigned to a brand new Utilized AI Engineering unit tasked with creating software-engineering puzzles for use as coaching information to enhance the coding abilities of Meta’s AI fashions. Many staffers derided the work as rote and boring in inner posts seen by Reuters. Meta informed Reuters that the corporate is starting to see progress with its redeployments, noting that information produced by the Utilized AI Engineering unit helped prepare an AI mannequin it launched final month.
Between the reassignments and the job cuts, headcount in some engineering items dropped as a lot as 30% by the tip of Could, in response to one doc and other people conversant in the planning.
Reuters additionally reported in April that Meta had mandated that monitoring software program be put in on U.S. workers’ units to seize their keystrokes and mouse clicks to show its AI brokers to copy how people work together with computer systems.
Believing they may be coaching their very own AI replacements, and irate over the dearth of element concerning the layoffs, many workers flooded Meta’s in-house communications community – known as Office – with offended screeds and gallows humor. Staffers replied to executives’ inner posts with photos of elephants, symbolizing that layoffs had been the elephant within the room, in response to examples seen by Reuters. Some workers straight clashed with Andrew Bosworth, Meta’s chief expertise officer, who oversaw the AI transformation and was defending it in posted feedback. One individual trolled Zuckerberg, sarcastically likening his inner announcement of an AI initiative for small companies to Prometheus, the determine from Greek mythology, giving hearth to humanity. Meta declined to make Bosworth obtainable for remark.
Employees morale sank. Meta’s inner measure of worker sentiment dropped from 74% favorable to 55% favorable, in response to the corporate’s half-year Pulse survey. Labor organizing efforts gained momentum. Amid the revolt, different inner information indicated that the tech on the coronary heart of the plan wasn’t delivering. Staff’ use of AI had resulted in an enormous improve within the code they generated, however with questionable impression on productiveness, in response to inner posts seen by Reuters. As an illustration, code adjustments made to the inner software program platforms and infrastructure workers use on the job had been up 220% year-over-year, in response to a publish by Bosworth in early June. However adjustments that led to new or upgraded options reaching Meta customers had been solely up 36%. As early as March, infrastructure groups had been additionally warning of “reliability warning indicators” brought on by the AI coding surge, in response to an inner publish. One other publish, in April, stated that unchecked AI brokers had been performing “large-scale, disruptive actions that people are unlikely to execute.” The outcome: Main technical and safety incidents, similar to service disruptions and attainable information leaks, spiked 40% from the earlier 12 months, with the time staffers needed to spend “firefighting” them up 70%, in response to the inner posts. The general public noticed a glimpse of the issue in early June after hackers exploited Meta’s new AI-powered buyer help bot to achieve entry to high-profile Instagram accounts, together with one for the dormant Obama White Home web page. Meta declined to touch upon the inner information about the AI disruptions.
With pressures piling up, Zuckerberg modified course. Hours earlier than Meta carried out its first wave of Undertaking OT layoffs on Could 20, he conferred once more along with his prime lieutenants and known as off the second wave of restructuring deliberate for November.
The following morning, Meta went forward with the ten% lower. However after the notifications went out, Zuckerberg posted a memo on Office telling Meta’s remaining workers he did “not anticipate different company-wide layoffs this 12 months” and expressed his need to offer them extra “stability.”
Executives then tried to shore up morale. They inspired unit heads to acknowledge workers’ emotions, paused the mouse-tracking program, let some workers within the new AI engineering unit switch again to their previous groups and dispatched Chief Monetary Officer Susan Li to spice up perks.
Within the weeks after the layoffs, a stream of empathetic govt posts arrived on Office, together with pledges to enhance snack high quality in workplace microkitchens and improve spending on journey and social occasions. Meta declined to make Li obtainable for remark for this report.
In early July, Zuckerberg made a shock look at an inner city corridor and conceded miscalculations on the reorganization’s timing. AI agent expertise, he stated, had not “accelerated” as shortly as he had anticipated. He added that he anticipated the tech to enhance and start exhibiting extra advantages within the subsequent three to 6 months.
As Zuckerberg has dialed again probably the most disruptive elements of the inner AI transformation, he has additionally launched a public-relations blitz positioning Meta as people-centric. The push included a video commercial proclaiming the corporate is “betting on individuals.”
Within the marketing campaign, which additionally included a latest 6,500-word essay outlining his imaginative and prescient for the way forward for AI, the CEO has highlighted Meta’s plans to make creating AI brokers extra person pleasant whereas portraying unnamed rivals because the true job killers. In an inner publish in June, he informed workers, “We’re the one main firm specializing in empowering individuals and placing the ability of this new expertise in billions of individuals’s fingers throughout all our merchandise – somewhat than primarily specializing in automating work.”
Nonetheless, Zuckerberg has caught to the phrases “company-wide” and “this 12 months” in discussing layoffs with workers, in response to his inner communications. That has prompted some workers to invest that he’ll proceed trimming the ranks by way of team-specific cuts or performance-based dismissals – or delay company-wide headcount reductions till subsequent 12 months.
The corporate is dealing with a money squeeze and investor scrutiny because of its dizzying AI spending, heightening the strain to seek out financial savings. Meta plans to speculate not less than $130 billion in AI chips and different infrastructure this 12 months, which analysts anticipate will eat up its working money for 2026, estimates from LSEG present.
In his essay on the way forward for AI, titled “The Future is for Everybody,” Zuckerberg predicts there could possibly be “an abundance of jobs sooner or later.” However some firms may find yourself with fewer workers. “Firm sizes could shrink – simply as they did within the transition from industrial giants to tech firms,” he writes. “However this does not imply fewer jobs general. It implies a bigger variety of firms with fewer individuals every.”
as a Dependable and Trusted Information Supply
