Six hundred General Motors employees got the same email on Monday, May 11. An ominous, oddly-timed 15-minute virtual meeting. A scripted message from HR. An abrupt end. “No appreciation or empathy. No questions. Nothing,” one data analyst told CNBC. The analyst had been at GM for more than a decade.
The reason GM gave is that it needs different skills now. “AI-native development,” to use the corporate phrase. Data engineering. Agent and model development. Prompt engineering. AI workflows. The company framed the cuts as a “transformation of the Information Technology organization to better position the company for the future.”
This is one of the cleanest examples we have, in public, of an enterprise telling its workforce that their skills are obsolete and that the only path forward is to fire them and hire different people. It is also one of the cleanest examples we have, in public, of corporate dishonesty.
Because the skills GM is firing for don’t exist anywhere yet. The entire stack is younger than GM’s last recall.
The Stack Didn’t Exist Six Months Ago
Let’s walk through what GM says it needs and when each of those things actually came into being.
Model Context Protocol (MCP) — the open standard for connecting AI systems to data and tools — was released by Anthropic in November 2024. It is eighteen months old as of this writing.
Microsoft Agent Framework — the orchestration layer enterprises are now using to build multi-agent systems — went general availability in 2025.
Prompt engineering as a job title started appearing in 2023. A “Senior Prompt Engineer” in 2026 has, at best, two and a half years of experience doing something that didn’t have a name when they started.
Agentic patterns, RAG pipelines, vector search at the application layer, SQL Server’s native vector type — all of it post-2022. Most of it post-2024.
There is no engineer alive with five years of experience building agentic AI systems, because the field is not five years old. Anyone GM brings in today with “AI-native” experience learned it by reading docs, building toy projects on weekends, and shipping something at work. That isn’t a credential. That’s a software developer doing what software developers have always done.
You know — the thing the people GM just fired do for a living.
I have been writing .NET on the Microsoft stack for thirty years. In the last eighteen months I picked up Microsoft Agent Framework, MCP server development, Semantic Kernel, and SQL Server 2025’s native vector storage. I built a multi-agent leadership system. I built an automated novel-writing pipeline running on local LLMs. I built an AI-powered semantic image search engine. I am not unusual. I am a competent engineer who paid attention. Every engineer GM walked out the door on Monday could have done the same.
GM did not give them the chance.
Companies Used to Hire for Attitude
There was a time — most of the twentieth century, in fact — when American companies hired for attitude and trained for skills.
Sidney Weinberg started at Goldman Sachs in 1907 as a janitor’s assistant. He was paid three dollars a week to brush hats and wipe mud from the partners’ shoes. He became a clerk, then a trader. In 1925 the firm bought him a seat on the New York Stock Exchange. In 1930 he became CEO. He held the job for thirty-nine years, until his death in 1969. The man had no degree. He had attitude, and Goldman taught him the rest.
David Geffen dropped out of three colleges and started in the mailroom of the William Morris Agency. He went on to build Geffen Records, sign Nirvana, John Lennon, and Aerosmith, and become one of the wealthiest men in American media. Mailroom to mogul.
Barry Diller did the same thing at William Morris. He left to become chairman and CEO of Paramount, and then Fox. The man who created the Fox television network started by sorting envelopes.
George Bodenheimer joined ESPN in the mailroom sixteen months after the network launched. Seventeen years later he was its president. He stayed in that role for thirteen years, the longest-serving boss in the company’s history.
Doug McMillon — the current CEO of Walmart — started as a teenager unloading delivery trucks in Tulsa. Karen Kaplan, CEO of Hill Holliday, joined as a receptionist at twenty-two, planning to save for law school. Chris Rondeau, CEO of Planet Fitness, joined as a receptionist at nineteen.
And then there is Mary Barra. The current CEO of General Motors started at GM at eighteen years old, as a co-op student inspecting hoods and fender panels at the Pontiac Motor Division. She was a teenager with a high school diploma and a willingness to learn. GM trained her. The company put her through what was then GMI — now Kettering University. It paid for her Stanford MBA. It moved her through plant management, then human resources, then product development. It taught her the business. It taught her every job she had until she was running the company.
Mary Barra is the most successful product of corporate training in the modern American auto industry. She is the living embodiment of the proposition that you hire smart, willing people and you teach them what they need to know.
And now she runs a company that just fired six hundred people because, it says, their skills don’t fit. Six hundred people whose skills almost certainly could have been made to fit, if GM had been willing to spend a fraction on training what it will now spend on severance, recruiting, relocation, and rehiring.
Mary Barra of all people knows training works. She is the case study.
What’s Actually Happening
If the “skills” justification is incoherent — and it is, because the skills GM is hiring for didn’t exist when half its IT staff was hired, and could be taught to those staff in a quarter — then something else is going on. There are three real mechanisms, and none of them are about skills.
One: comp reset. A senior engineer in Warren, Michigan, with fifteen years of tenure, full benefits, stock vesting, and an institutional-knowledge premium is expensive. Fire that engineer with two months’ severance and rehire someone at a lower band under an “AI Engineer” title and the salary delta compounds across years. Severance is one-time. The lower salary is forever.
Two: geographic arbitrage. Austin and Warren are not cheap labor markets. When GM announces that it is “transforming” its IT organization and then quietly fills the new headcount through Accenture, Tata, Wipro, or any of the other usual suspects — frequently with engineers based offshore — the cost structure changes dramatically. The “AI skills” line is the political cover for what would otherwise look like ordinary offshoring. You can’t fire a thousand Americans and replace them with contractors in Hyderabad without a story. “We need AI talent” is the story.
Three: Wall Street theater. GM stock trades at a forward P/E of about six — roughly half of Ford’s, and a tiny fraction of any company that calls itself “tech.” There is real money in re-rating GM as a software-defined-vehicle company rather than a cyclical manufacturer. The IT layoff is part of the costume. “We are firing legacy workers to hire AI talent” reads to Wall Street as “we are not a 1970s industrial company anymore.” Whether it’s true matters less than whether it’s said.
None of these have anything to do with the skill profile of the engineers being walked out the door.
The Domain Knowledge GM Is Walking Out the Door
Here is what GM is actually losing.
Auto manufacturers run on systems that are older than most of their engineers. Dealer management platforms built on the AS/400. Parts databases with thirty years of accumulated SKU mappings. Financial systems integrated with GM Financial’s lending platform. Regulatory reporting pipelines for NHTSA, EPA, CARB, and dozens of state DMVs. Manufacturing execution systems on the plant floor that connect to forty-year-old PLCs. Warranty claims processing. Recall workflow management. Telematics ingestion from OnStar. Connected-vehicle data pipelines. Supplier portal integrations reaching down into the bottom of the Tier 2 and Tier 3 chains.
A senior engineer in Warren who has been there for a decade knows where the bodies are buried in all of it. They know which integrations are fragile, which batch jobs run at 3 a.m. on Tuesdays, which systems have undocumented dependencies on a SQL Server instance that nobody remembers buying.
That engineer is precisely the person you would want learning agent frameworks and building MCP servers — because the value of an enterprise AI agent is not in the model. The model is a commodity. The value is in connecting the model to the company’s actual data and systems, safely, with the right access controls, in the right context, with awareness of the business rules that make this company different from the next one. That work requires institutional knowledge.
The “AI-native engineer” GM hires from outside spends the first six months learning what the fired engineer already knew. Often longer. Sometimes they never get there, because the fired engineer was the documentation.
This pattern is not subtle, and it does not surprise anyone who has actually done enterprise software. The Detroit News quoted Sam Abuelsamid of Telemetry Agency observing that companies cut IT for AI and then quietly rehire because the AI doesn’t work the way executives think it does. “It doesn’t always work the way they expect it to,” he said, “so they have to hire people back, or they have to have people that are double-checking the work that AI is doing.”
That is the future GM is buying with this round of cuts. They will be rehiring some of these people inside of eighteen months. Probably the same people, through a contracting firm, at a higher daily rate.
Now Let’s Talk About the Product
There is one more thing that needs to be said about a company telling its engineers their skills are inadequate, and it is this: General Motors is not exactly in a position to throw that stone.
In April 2025, GM recalled 597,630 trucks and SUVs — Silverados, Sierras, Tahoes, Suburbans, Yukons, Escalades — because the 6.2-liter L87 V8 engines could catastrophically fail. The connecting rods and crankshafts had manufacturing defects. The engine could seize while you were driving. At least twelve crashes and twelve injuries had already occurred. GM had received twenty-eight thousand field complaints. Owners had been complaining since 2021. GM closed three internal investigations into the problem and only acted after NHTSA opened a federal probe.
GM’s solution for vehicles that pass inspection? Thicker oil. They are recommending dealers switch from 0W-20 to 0W-40 and hope for the best. On an eighty-thousand-dollar Escalade.
It gets worse. One owner filed a class-action lawsuit for more than five million dollars in damages because the replacement engine GM installed in his 2023 Yukon Denali — to fix the recall — had the same defect as the original. GM is selling replacement engines from the same defective supplier. NHTSA has now opened a fourth investigation, because owners are reporting that the recall fix didn’t work.
Want to talk about the EVs? The Chevrolet Bolt — GM’s mainstream electric vehicle — had to be recalled across every model year of its production from 2016 onward. Eighteen Bolts caught fire while parked. GM ultimately recalled every Bolt ever built and replaced the batteries. LG paid GM 1.9 billion dollars in a settlement because the batteries were defective from the supplier. GM’s own quality assurance had failed to catch a defect across a six-year production run on its flagship EV.
The Ultium platform, GM’s grand strategic bet on electric vehicles, was supposed to deliver 400,000 vehicles in North America by mid-2024. Actual 2023 deliveries: 13,838. Two-thirds of those were one model, the Cadillac Lyriq. By October 2024 the company had abandoned the 400,000 number altogether. They are now retiring the Ultium brand entirely.
Want to talk about Cruise? GM bought a controlling stake in the robotaxi unit for 581 million dollars in 2016. By December 2024, when GM shut Cruise down, the unit had accumulated more than ten billion dollars in operating losses and generated under five hundred million in revenue. Along the way, a Cruise vehicle dragged a pedestrian roughly twenty feet down a San Francisco street. Cruise then misled California regulators about it. The company’s autonomous-vehicle license was suspended. The co-founder resigned. The unit was wound down. Ten billion dollars on the fire.
Want to go back further? The ignition-switch scandal. Mary Barra became CEO of GM in January 2014. Within weeks, the company began recalling 2.6 million vehicles because a defective ignition switch could shut off the engine — disabling power steering, brakes, and airbags — if a heavy keychain bumped it. The defect was linked to at least thirteen deaths in the official tally; some independent analyses put the real number above one hundred. An internal investigation found an eleven-year “history of failures” inside GM. A senior senator described “a culture of cover-up.” An engineer changed the part design without changing the part number — a deliberate act to mask the existence of the defect from anyone trying to track it. Fifteen executives were fired. GM paid a 900-million-dollar fine.
Mary Barra, to her credit, was direct about it at the time. “We had a cost culture,” she said in 2014. “Now we have a customer culture.”
Twelve years later, GM is firing six hundred people because they are too expensive, while shipping engines that seize, EVs that have caught fire, a robotaxi that dragged pedestrians, and trucks whose replacement engines have the same defect as the originals.
That doesn’t sound like a culture problem in the IT department.
Where the Skills Gap Actually Lives
GM has new tech leadership. Sterling Anderson, the chief product officer, co-founded Aurora and is a veteran of the autonomous-vehicle industry. Behrad Toghi came in from Apple as AI lead. Rashed Haq, the new vice president of autonomous vehicles, spent five years at Cruise — the company GM bought for half a billion dollars, ran into the ground for ten billion dollars, and then shuttered. Haq was Cruise’s head of AI and robotics during exactly that period.
So the executive team responsible for declaring that the IT department’s skills no longer fit includes a man whose previous job was head of AI at the project GM just admitted was a ten-billion-dollar mistake.
I don’t begrudge Haq the new role. People learn from failures — that’s the entire point. The willingness to take someone who has spent five years inside a colossal disaster and put them in charge of the next attempt is, ironically, an act of training-on-the-job belief. GM extends to its executive suite exactly the patience and faith in human development that it just refused to extend to its engineers.
The engineers being fired this week didn’t run Cruise into the ground. They didn’t ship Bolts that caught fire. They didn’t design ignition switches that killed people. They didn’t promise four hundred thousand Ultium vehicles and deliver fourteen thousand. They are paying the cost for management’s failures, packaged as a story about their skills.
The Mailroom Is Locked
Sidney Weinberg, the janitor’s assistant who ran Goldman Sachs for thirty-nine years, would not be hired by Goldman Sachs today. He had no degree. He had no internship pipeline. He had no LinkedIn profile. He had attitude and a willingness to learn, and Goldman gave him a job sweeping the floor.
Mary Barra would not be hired by GM today as an eighteen-year-old co-op student inspecting fenders, because GM has spent the last thirty years optimizing the path between teenager and CEO out of existence. The mailroom is locked. The door has “AI-native” written on it.
Companies that hire for fit and train for skill build careers. Companies that fire for skill and rehire for fit build resumes. Six hundred resumes hit LinkedIn on Monday because GM decided that was the cheaper option this quarter.
It will not be the cheaper option in eighteen months, when the new “AI engineers” need someone to explain how the dealer management system actually works and there is no one in the building who knows. It will not be the cheaper option in three years, when the next round of leadership realizes that institutional knowledge is the only durable advantage left in an industry where the technology stack is commoditized by Tuesday.
But by then, Mary Barra will probably be retired, the strategy will be someone else’s problem, and the next CEO — almost certainly hired in from the outside — will explain that the company needs different skills now.
The mailroom worked because companies believed in the people in it. GM doesn’t believe in its people anymore. It believes in its press releases.
And it shows, on every dealer lot in America.
Gal Ratner is CTO of WhiteStar Labs and Chief Architect at Prana Entertainment. He has been building enterprise software on the Microsoft stack for thirty years and has been shipping agentic AI systems since the field had a name.
The Theory of Compensation as Regards the
Workpeople Displaced by Machinery