The Better You Get at AI, the Closer You Get to Replacing Yourself

What do you call it when a company spends five years telling people to work from wherever they'd like, then hands a large share of them a moving date. Uber calls it "establishing clearer principles for where roles and teams should sit." It's one version of a pattern running through this week's news: the promise and the number attached to it keep drifting apart, whether that promise is about jobs, flexibility, fairness, or reward.

Employer branding works better when people share what they know. EBClub brings practitioners together to exchange ideas, compare experiences, and learn from the people doing the work.

Explore EBClub

The clearest case is the newest one. Data centres in the UK are on track to create about a quarter of the jobs the tech sector said they would, according to new thinktank analysis, a gap wide enough that it's worth asking what else this week's numbers don't quite add up to.

This Week’s Paper Cut's

Watch or listen here ↯

We keep tracking to a minimum. The video is hosted on YouTube and the podcast is hosted on PodBean, which may place cookies that measure viewing and personalize content. We only load the player after you consent to these cookies. You can update your cookie settings any time.

The jobs AI data centres promised keep shrinking once someone checks the number

Environmental thinktank Verdant this week estimated that all currently planned UK data centres will directly employ around 10,400 workers, against the more than 40,000 cited in projections from industry lobby group TechUK. The data centres already built have created fewer than 4,400 jobs between them, Verdant found, about 8.6 jobs for every megawatt of energy they consume, against the 43.7 jobs per megawatt claimed in TechUK's own projections. For comparison, Verdant put the figure at roughly 400 jobs per megawatt-equivalent for a car plant and over 3,600 for a hospital. James Meadway, Verdant's co-director, said the UK is "being rushed into supporting a frantic pace of datacentre buildout on very flimsy economic grounds." TechUK disputed the methodology, standing by its estimate of 43,500 jobs currently supported and 40,000 more by 2035.

The jobs argument is not getting stronger as facilities age, either. Meta is reportedly testing robotic arms from firms including ABB, Kinova and Watney Robotics to automate data centre maintenance work, tasks like hot-swapping network cables and cycling power to server racks, according to Ars Technica's reporting. One Meta staffer said a single Kinova arm, if the trial succeeds, could replace up to 80% of the facility's workload. "We thought those of us performing the physical tasks were safe for a while, but not anymore," the worker said. "It's coming for us all, unfortunately."

None of this is slowing the investment case. Oracle told investors this week that its data centre revenue is accelerating as its AI infrastructure push gains ground, the Financial Times reported, even as the employment case for building the facilities keeps getting smaller. That is not a contradiction so much as the point: the money case for data centres and the jobs case for them were never really the same argument, and this week is the clearest evidence yet that only one of them is holding up.

Employer Brand POV

Measure the jobs a project or a programme actually creates against the number used to win the argument for it in the first place, whether that is a new facility, a reorg, or an AI rollout. Verdant's analysis exists because TechUK's own projections were checkable against public planning data, and if your organisation is making a jobs or growth claim to a community, a government or your own workforce, assume someone will eventually check it the same way.

Companies are quietly grading how enthusiastic you sound about AI, not just whether it works

Duncan Trevithick's employer runs a bonus scheme tied to AI use, and he's not sure the maths works in his favour. "I do not receive two days off or a 40% pay rise. The higher output simply becomes the new baseline," he told the BBC. "In the short term, that may help me get promoted. In the longer term, I have helped prove how much of my job no longer requires me." He's not alone in noticing the shift. Accenture chief executive Julie Sweet has said plainly that using AI is now how the firm operates, full stop, and Disney, Meta, JPMorgan and KPMG have all reportedly introduced internal "AI leaderboards" tracking staff usage. Coinbase fired engineers who skipped mandatory AI training. The pressure is landing despite thin proof that any of it pays off: McKinsey found 94% of companies have yet to see significant value from their AI investment.

Job seekers are absorbing the same signal from the other side of the table. Felisha Case, laid off from a LinkedIn programme manager role in May, said she's now routinely asked in interviews how she used AI in her old job, despite limited experience to draw on. "It can feel like we're being asked to go back and get a whole new education in AI," she told Business Insider. A separate Big Tech employee interviewing elsewhere admitted he fakes enthusiasm he doesn't feel: "if it all magically disappeared tomorrow I wouldn't shed a tear," while applying for AI roles anyway because, in his words, that's "where the money is." Meanwhile a graphic designer, Danny Williams, told the BBC he's simply losing work to small businesses generating their own posters with AI tools, a rare example this week of the job loss being real rather than performative.

Not every employer is leaning the same way. Duolingo's chief executive said in April the firm had dropped AI use as a performance-review factor after staff started asking whether they were meant to use it "for AI's sake." Amazon reportedly shut down an internal AI leaderboard after employees were caught gaming it with needless tasks. And KPMG's incoming talent chief, Jason LaRue, argued the whole premise of treating this as a phase to get through is wrong: "this is not a sort of a Y2K kind of moment, where it's going to sort of come and go... it'll probably be the slowest and least amount of change we'll ever go through going forward."

Employer Brand POV

Read what you're actually rewarding before you build a leaderboard or a bonus scheme around AI use. If the metric is visible usage rather than judgement or better decisions, you will get visible usage, employees performing fluency for a dashboard, and Duolingo and Amazon have both already had to walk that back. Decide what "good" looks like first, then measure that, not the easiest thing to count.

After the cuts: some managers are coming back, and so are some of the people who left

Meta is asking some individual contributors in its Applied AI division to move back into management, a person familiar with the matter told Business Insider, an opt-in reshuffle inside the roughly 7,000-person unit Meta launched this year. It's a partial reversal for a company that spent much of the past year cutting middle management and promoting flatter teams, with Meta's own layoffs earlier in 2026 having heavily targeted managers specifically. Not every company is moving the same direction at the same time. Uber, by contrast, announced this month it was cutting half of its "micro-teams" as part of its own restructuring.

Uber's cuts came with a second announcement that changes more daily lives than the layoff itself. Of the roughly 30,000 employees remaining after the 3,300 job cuts, fully remote roles will now be capped at 1% of the workforce globally, a ceiling rather than a target, down from a share built up over five years of hiring people wherever they lived. Everyone else is being told to move to an office matched to their team, following new hub rules: global teams to New York or San Francisco, regional teams to designated hubs, engineering to tech hubs. The existing three-day-a-week hybrid policy stays, and Uber says it will now actively count compliance rather than leave it to managers. Chief executive Dara Khosrowshahi framed the reversal around a lesson from the pandemic, that people solve problems better sitting together, though the timing lines up with cutting headcount seven or more layers below the CEO by 20% and halving the number of teams with only one or two direct reports.

Against that backdrop, some companies are trying to win back the people they, or someone else, let go. Deel says rehires at the company have more than doubled between 2024 and 2026, and nationally, 3.4% of new US hires by the end of 2025 were "boomerang" employees returning to a former employer, up from 3.1% two years earlier, according to Revelio Labs. Deel chief executive Alex Bouaziz says former high performers who left on good terms get fast-tracked: one employee texted him a selfie asking to come back and was talking to a recruiter within 48 hours. "We tend to focus on the highest performers," Bouaziz said, adding that former employees who ask and don't get a yes still leave with a more positive view of the company for having asked.

Employer Brand POV

Locate, literally, where your remaining people are supposed to be, and whether a return-to-office order matches where the work actually happens. Uber's policy ties attendance to specific buildings for specific teams, not a blanket "come back" memo, and that specificity is what makes it enforceable rather than symbolic. If you're simultaneously cutting managers in one part of the business and asking for them back in another, as Meta is, say so plainly rather than letting people read it as inconsistency.

When a company gets it wrong, the real question is who actually pays for it

Changzhou Xingyu Automotive Lighting Systems, a Chinese lighting supplier to Volkswagen, Mercedes-Benz and BMW, fired 107 of the 372 fresh graduates it had hired this year, barely a month into their jobs, before backlash forced a public reckoning. Local authorities found the company had handled the terminations in a "simplistic and blunt" manner. Xingyu's response went further than an apology: general manager Zhou Xiaoping will lose a year's salary, deputy general manager Li Shujun will lose six months' salary and have his duties adjusted, the HR director was dismissed outright, and the HR department head was demoted. Each affected graduate received a 15,000 yuan lump sum, with six months' wages promised to anyone still unemployed after three months. As of this week, 71 of the 107 had already found new jobs. Volkswagen, Mercedes-Benz and BMW have all opened their own reviews of their supplier over the incident.

Ireland offers a quieter version of the same question, accountability without the scandal forcing it. Only 20% of Irish employers now believe the EU's pay transparency directive will have a positive impact on their workplace, down from 31% just four months ago, according to research for law firm Mason Hayes and Curran. More than half say they don't currently disclose salaries in job ads and aren't comfortable with staff discussing pay with colleagues, even as the directive's phased rollout continues into next year. "The legislative timeline in Ireland has slowed, but the compliance requirements haven't changed," said the firm's Catherine O'Flynn.

And a University College Dublin-led survey of 4,300 workers across Ireland found that AI's productivity gains are landing almost entirely with employers rather than staff: most workers who save time using AI reinvest it into more work rather than a better work-life balance, a third report their pace of work intensifying, and just 4.7% report any increase in earnings as a result. Ernst & Young offered a rare counterexample this week, announcing $100 million in US bonuses specifically for staff who demonstrate skills like adaptability, judgement and AI experimentation, rewarding the human contribution rather than just the AI usage metric.

Employer Brand POV

Credit the specific behaviour you actually want repeated, whether that's a manager's judgement call during a hiring mistake or genuine skill working alongside AI, rather than the number that happens to be easiest to track. Xingyu was forced into specific, named accountability by public pressure. EY chose to reward specific, named behaviour without needing the pressure first. Both are more credible than a vague statement, and the gap between them is entirely about whether you act before or after someone makes you.

The Cut List

Volkswagen's restructuring will cost an estimated €16 billion ($18.6 billion), a person familiar with the matter told Reuters. The plan includes exploring alternatives for four German plants, with phasing out production at Emden and Zwickau estimated to cost about €1 billion each, and Neckarsulm and Hanover about €2 billion each, alongside roughly €10 billion set aside for cutting up to 60,000 jobs worldwide. Separately, the US added 162,000 jobs in August, about three times the expected 55,000, led by leisure, hospitality, construction and healthcare, according to the Bureau of Labor Statistics. The only two major industries to lose jobs were both white-collar fields, and the information sector, which includes much of tech and media, remains about 12% below its 2022 peak. Tech occupation employment still grew by 86,000 in August despite 14,700 tech job cuts, according to a CompTIA analysis of the same data.

Takeaways

• UK data centres already built have created fewer than 4,400 jobs, about 8.6 per megawatt of power used, against an industry claim of 43.7, according to thinktank Verdant.
• Meta is testing robotic arms to automate data centre maintenance tasks; one arm could reportedly replace up to 80% of a facility's workload.
• A Gi Group poll of 1,881 UK jobseekers found 75% would not be put off applying to a company that assessed AI proficiency in performance reviews, even though McKinsey found 94% of companies have yet to see significant value from AI.
• Duolingo and Amazon have both rolled back internal AI-usage tracking after employees were found gaming the metrics.
• Uber's remaining roughly 30,000 employees face a global cap of 1% on fully remote work and reinforced three-day office attendance, alongside the 3,300 job cuts announced this month.
• Meta is asking some employees in its Applied AI division to return to management roles, a partial reversal after a year spent cutting managers and flattening teams.
• A Chinese auto supplier fired 107 new graduates about a month into their jobs, then compensated them and disciplined four managers, including a year's lost salary for its general manager, after public backlash.
• Only 4.7% of Irish workers who save time using AI report any increase in earnings from it, according to a University College Dublin-led survey of 4,300 workers.

SOURCES