AI Will Take Your Job Someday, Says Company That Did It on Tuesday
A short story of doomerism, ignorance, and poetry. At least the coffee is good, unlike the arguments I'm reading about the future.
The Missing Subject
Somewhere inside Oracle's 2026 annual filing, a sentence is trying very hard to look innocent.
"The adoption and deployment of AI technologies across our operations have resulted, and may continue to result, in reductions to our workforce." Oracle disclosed that its headcount had fallen from 162,000 to 141,000 over the fiscal year, a 21,000-person reduction. The sentence arrived in a regulatory filing, which means it had been scrubbed, reviewed, blessed, and dressed for the SEC before being allowed outside.
It is a beautiful little corporate fossil. No manager appears. No board appears. No executive appears. No one chooses. No one signs. No one fires. AI technologies are adopted and deployed, and then reductions simply result, like condensation on a window.
This is where the grammar gives the game away.
For most of spring, the public argument was still staged in the future tense. Will AI take our jobs someday? Five years from now? Ten? Somewhere around 2035, after the robots finish graduate school and discover LinkedIn? The AI is coming for your job, then your wife, and, in the more ambitious threads, your entire life. The panic industry stood on a chair, screaming about a tsunami on the horizon. The corporate euphemism industry stood ankle-deep in water, quietly renaming it "strategic alignment."
Meanwhile, the present tense was having a perfectly active little career.
I first noticed it in May, between the scheduled panics. Underneath the doom scroll, smaller headlines kept sliding past about companies already reorganizing actual jobs around AI. Present tense, no cinema. It did not look like an evil robot snatching people from their desks. It looked like what new tools have always done to work. Humanity has fought this beast before, and the fight never produces winners or losers. One generation panics for a few years, and the next one continues its life in oblivion, living inside the change without ever noticing it was new. So I wanted to see if I was right, and I set up an automated weekly tracker to collect every layoff announcement that arrived dressed in AI language, and let it run.
For six weeks, the tracker kept collecting the same object in different costumes. Layoffs attached to AI strategy. Workforce reductions bundled with record revenue. Cuts announced beside buybacks, model partnerships, agentic roadmaps, restructuring charges, and shareholder-friendly phrases that smelled faintly of bleach. By mid-June, tracker estimates were already ranging into the hundreds of events and roughly 152,000 to 184,000 affected workers, depending on the source. Challenger had reportedly put AI among the most-cited layoff reasons for the third straight month.
The annoying part is that nobody in the official language ever seems to have done it.
That is the missing person in the AI layoff sentence. The subject.
The Weather Report
The first family of euphemisms is meteorological. An era arrives. The company adapts. The atmosphere changes. Executives become humble umbrella-holders in a storm nobody ordered.
Amdocs cut up to around 3,000 people while explaining that it was "adapting work processes to the AI era." Wix moved roughly 1,000 people toward the exit while preparing to "adapt to the new automation era." GitLab gave us "Act 2" and the "agentic AI era," which made a 14% reduction sound like narrative structure. Cloudflare, after record revenue, framed its cuts around how a "world-class, high-growth company operates and creates value in the agentic AI era."
The era is doing incredible work here. It arrives like weather over Vestfold. Management is just standing there in a nice quarter-zip, watching history happen to the payroll system.
This language is useful because weather has no defendant. Nobody sues a cold front. Nobody asks a fog bank why it removed the QA team. "The AI era" sounds like a condition of existence, a historical pressure system, a force too large for human fingerprints. The company did not decide to cut. The company adapted. Survival posture. Evolutionary squint. Very brave. Very humid.
The panic industry loves the same trick from the other side. It also talks about AI as weather. The wave is coming. The flood is coming. The replacement event is coming. Always coming. Always future. Always cinematic enough to sell a keynote, a think piece, a consultancy deck, or an emergency webinar with a countdown clock.
Both sides benefit when the sentence has no subject. The doomers keep the drama parked safely in tomorrow. The companies keep the decision hidden inside "the era." The rest of us are asked to debate whether the water will arrive someday while the carpet is already making that wet little noise.
The Taxonomy Game
The second family is classification. If you can invent a category, you can make the firing look like sorting.
Cloudflare gave the month one of its most useful specimens with "measurers." Middle management, finance, legal, audit, revenue recognition. The "vast majority" of those laid off, according to the tracker, were described as measurers rather than doers. That is not just a label. That is a tiny moral trial conducted inside an org chart.
Once the category exists, the conclusion starts pretending it was always obvious. Doers build. Measurers measure. Builders are the future. Measurers, unfortunately, have been found guilty of standing near spreadsheets during the agentic age.
Accenture gave the taxonomy a more spiritual costume. The company described "upskilling our reinventors" while exiting people for whom "reskilling is not a viable path." That phrase deserves a small, cursed museum plaque. Reinventors survive. The others are not simply laid off. They become linguistically un-reinventable. A business decision gets dressed as a test of metaphysical adaptability.
Coinbase had "pure managers" versus "player-coaches." The sports metaphor is doing laundry at high heat. A pure manager sounds soft, ornamental, overpaid. A player-coach sounds lean, sweaty, heroic, probably allergic to meetings. Under that vocabulary, reducing management layers becomes a character upgrade for whoever remains.
Standard Chartered, bless the accidentally honest, reportedly reached for "lower-value human capital." Investor Day language is always where the mask slips because the room is full of people trained to clap when humanity becomes a portfolio input. "Lower-value human capital" is what happens when the spreadsheet achieves sentience and immediately joins a golf club.
The taxonomy game has one job. It tells the fired what they were, and it tells the survivors what they had better become.
The Geometry Lesson
Then comes shape language, which is adorable because shapes cannot testify.
Cognizant had "Project Leap" and the "broader and shorter pyramid." GitLab flattened management layers. Other companies rebalanced, reduced complexity, shortened hierarchy, streamlined, aligned, and optimized. The human event becomes a diagram with arrows. A pyramid suffers. A layer disappears. A structure becomes elegant.
There is a reason this works. Geometry has no crying spouse, no cancelled mortgage plan, no laptop lockout at 7 a.m., no manager reading from a script with the emotional range of a printer jam. Geometry is clean. It offers distance. It lets executives discuss layoffs as if they are renovating a kitchen.
The phrase "broader and shorter pyramid" is especially special. It makes thousands of people sound like an architectural inconvenience. Nobody says the quiet part, which is that a shorter pyramid requires removing stones. Possibly stones with rent.
This is the great magic trick of corporate geometry. It translates agency into design. It takes "we cut these people" and turns it into "the shape changed." The shape changed because someone changed it, obviously, but the sentence has already fled the scene.
The Outsourced Verb
British American Tobacco added another species to the field guide with "transferred to third-party strategic partners." Around 9,000 roles were affected under its "Fit2Win" transformation programme, with roughly 5,500 cut outright and another 3,500 moved into that clean little outsourcing phrase.
Transferred is doing charity work here. Strategic partners are doing the same thing, except wearing a lanyard. A person is no longer employed by the company, but the sentence refuses to touch the hot stove. It slides the verb away from fired and toward logistics. Almost travel-like. Your job has been transferred to a third-party strategic partner. Please collect your dignity from carousel four.
Programme names deserve their own ashtray. Fit2Win. Project Leap. Act 2. These are not explanations. They are packaging. The cut inherits a logo and suddenly looks like part of a campaign. It has branding, therefore it must have strategy. It has capitalization, therefore it cannot be panic.
This is how the corpse gets a name badge.
The named programme also gives everyone inside the company a safe phrase to repeat. Nobody has to say "the layoffs." They can say Fit2Win, and the room can move on. Language becomes a corridor around the body.
The Denial Wing
The denial cases are almost more interesting than the confessions.
Intuit became a clean specimen because the CEO reportedly said the company's 17% reduction had "nothing to do with AI," while the same press cycle involved AI partnerships with OpenAI and Anthropic and language about reducing complexity and coordination-heavy roles. Uber denied AI linkage in its own people-division cuts. Expeditors praised AI deployments for productivity in May, cut around 230 tech workers in June, and gave no stated reason for ending a decades-long no-layoff policy.
This does not prove every denied cut was secretly caused by AI. That would be lazy, and lazy is how you end up with a podcast panel.
The denial matters because it shows the linkage has become radioactive. Companies want the investor glow of AI adoption, but they do not always want the worker-facing stink of AI replacement. They want to be seen as disciplined, modern, agentic, future-ready, operationally lean, and extremely serious about productivity. They do not want the headline to read: "Local Humans Sacrificed To Spreadsheet God Wearing Chatbot Mask."
So we get a strange little dance. AI is central to the strategy. AI is transforming the operating model. AI is changing how work gets done. AI is driving productivity. AI is everywhere except inside the layoff sentence, where it suddenly becomes shy and asks not to be named.
The denial becomes its own data point. It tells us where the bruise is.
The Eyebrow
Here is where the eyebrow goes up and refuses to come down.
So many of these cuts are happening around strength, not collapse. Cisco announced fewer than 4,000 job cuts on the same day it reported record quarterly revenue of $15.8 billion and described the need to shift investment toward AI-era demand. Cloudflare cut about 20% after record revenue. GitLab announced a 14% reduction alongside 23% year-over-year revenue growth. BILL cut roughly 30% and announced a $1 billion buyback in the same broad earnings context. Groupon's stock rose after its AI-native pivot and cuts.
This is the piece of the story that makes the old layoff script look dusty. The classic version was distress. Revenues down, demand weak, costs too high, sad piano music in the all-hands. The 2026 AI version often sounds different. Revenue is fine. Sometimes better than fine. The company is not starving. It is reallocating.
That word deserves suspicion. Reallocation is what you say when payroll becomes raw material. Salaries are not disappearing. They are changing state. Human beings go out one door, AI infrastructure orders come in through another, and everyone in investor relations pretends this is physics.
The market has learned to read the layoff as a proof-of-seriousness signal. A company says AI is the future. Fine. Cute. Everyone says that. A company cuts staff while saying AI is the future, and suddenly the story has teeth. The cut becomes evidence that management is disciplined enough to make the future hurt.
ServiceNow completed the loop beautifully. A company that sells AI efficiency confirmed hundreds of cuts while hailing "real AI efficiencies" inside its own business. That is the dogfooding layoff. The workforce reduction becomes a product testimonial. Look, our platform works. We know because fewer of us are here.
Somewhere, a sales deck felt warm.
The Control Group
Cisco matters because Cisco ruins the alibi.
The useful thing about a control group is that it shows what was possible. In the tracker, Cisco stood out because it was comparatively blunt about automation replacing workers in the broader AI restructuring cycle. Later, Cisco's own May memo still used plenty of polished language, but it also plainly connected workforce reduction to shifting investment, cost structure, and AI-era demand. The company said the reduction would affect fewer than 4,000 jobs and described strategic investments in silicon, optics, security, and employees' use of AI.
That matters because it collapses the most boring defense in the room: this is just how corporate communications works.
No. It is how corporate communications works when corporate communications wants the subject to disappear.
Cisco proves the softening is optional. A company can say the thing more directly. It can say there are reductions. It can say the reductions are linked to where investment is moving. It can say AI is part of that movement. It can put the decision somewhere near the decision-makers.
Everyone else chose the weather report.
That does not make Cisco morally pure. Please. We are not children and this is not a sticker chart. It makes Cisco structurally useful. The existence of a less euphemistic sentence demonstrates that the more euphemistic sentences were choices.
Why the Sentence Looks Like That
The audience for a layoff announcement was never primarily the people being fired.
Those people are the occasion. They are not the market.
The real audience is investors, courts, regulators, journalists, future plaintiffs, and the survivors who have to keep producing after watching a meeting invite turn into a trapdoor. Every sentence has to do several jobs at once. Reassure shareholders. Reduce liability. Preserve morale. Signal strategic seriousness. Keep the remaining staff frightened in a productive direction. Avoid giving critics a clean quote to beat you with. Avoid saying "we replaced people with AI" while making sure Wall Street hears "we are serious about AI."
Passive voice is legal armor. Era-language is liability management. Taxonomies are morale management. Programme names are brand management. Denial is reputation management. Geometry is grief management for people who cannot say grief.
This is why the same vocabulary now travels so easily across industries. In July, the tracker's in-window cases included a tobacco giant, a carmaker, and a networking company. British American Tobacco had Fit2Win and strategic partners. Volkswagen ran the older playbook: a plan for roughly 100,000 cuts that leaked through Manager Magazin instead of arriving in a press release, got blocked by the unions inside two weeks, and never needed AI language at all, because "Chinese competition" and "EV-transition costs" are the respectable fogs that predate the chatbot. Cisco had opportunity and cost structure. The AI transformation dialect had moved out of Silicon Valley and into legacy industry, and where it hadn't arrived yet, older fog was already on tap.
When a SaaS company, a bank, a cigarette company, and a carmaker all reach for the same fog machine, you can stop pretending the fog has anything to do with technology. It has been standardized. Procurement has a vendor for it. The words arrive pre-approved and posture-tested, ready to survive a deposition, and they describe the actual layoffs about as well as a scented candle describes a house fire.
And the acceptable posture is this: nobody fired anyone. The future happened nearby.
The Era Can Apparently Un-Arrive
Now comes the funniest part, because corporate history has a slapstick rhythm if you are dead inside in the correct way.
The walk-back cycle is already running.
Meta's Zuckerberg reportedly admitted "mistakes" in the company's AI restructuring after roughly 8,000 cuts and thousands of reassignments toward AI work. The tracker notes the important distinction: the execution was conceded, not the premise. The idea that AI justified the restructuring remains safely outside the apology zone.
Then came the broader reversal narrative. By early July, coverage and survey data were already pointing at leaders who had cut staff for AI and later called it a mistake. The tracker cites an Orgvue survey where 55% said so, a finding that had been sitting in public since spring 2025, waiting for July's coverage to make it fashionable, and Robert Half finding 32% had rehired the same or similar roles. Ford rehired engineers. Commonwealth Bank reinstated customer-service staff after an AI voicebot problem. IBM talked about expanding entry-level hiring.
This is why the language matters.
Every phrase in the field guide is pre-loaded against the reversal. If the AI era required the cut in May, what exactly happened in July? Did the era apologize? Did the automation front move offshore? Did the broader and shorter pyramid request planning permission for an extension? Did the lower-value human capital become mysteriously higher-value once the chatbot started chewing through edge cases like a raccoon in a server room?
The future tense gave everyone cover. The euphemism gave everyone distance. Together, they let the present happen quietly.
That is the receipt collection. Not proof that every layoff was purely caused by AI. Not proof that every executive is a cartoon villain stroking a GPU under the desk. The evidence is sharper than that. It shows a repeatable grammar. It shows the same missing subject, over and over, across sectors, filings, memos, and investor calls.
The panic industry got the tense wrong because tomorrow is easier to sell.
The euphemism industry got the subject wrong because responsibility is expensive.
The tracker's accusation is simpler, and therefore ruder.
The jobs were not waiting in the future to be taken. They were already being removed in the present, one passive sentence at a time. And when the reversal comes, as it already has begun to, the same companies will discover that the era was never weather.
It was management.
Funny how quickly history learns to use active voice when it needs the people back.