AI becomes the top reason US employers cut jobs, and the cuts spread past tech

For four straight months, artificial intelligence has been the single most-cited reason US employers give for cutting jobs, a streak with no precedent in the records Challenger, Gray & Christmas has kept since 2023. In May, companies tied 40% of the roughly 97,000 announced cuts to it. Microsoft's roughly 4,800 layoffs this month are part of that pattern, and Amazon and Meta have each let thousands go this year on the same logic.

The companies rarely say AI took the specific jobs. Microsoft's chief people officer told staff the eliminated roles are "not being replaced by AI," then added that "AI is changing how work gets done." Andy Challenger of the same firm put the mechanism more plainly: whether or not a given job goes to a model, the budget that funded it does. Amazon, Alphabet, Meta, and Microsoft have together guided about $700 billion in capital spending this year, most of it on AI data centers and chips, and payroll savings help pay that bill.

The cuts are also leaving tech. Citigroup is targeting around 20,000 job eliminations by the end of 2026, with its outgoing finance chief tying the continued decline to AI tools, and PayPal has announced roughly 4,760 cuts over the next two to three years. Transportation is now the second-largest source of announced layoffs in the country; the freight broker C.H. Robinson has cut its workforce by 31% since 2022 under a strategy it calls "Lean AI."

A jewellery store owner in Michigan used an AI tool to set repair prices across her 50 locations. A car dealer in Germany used the same kind of tool to replace his spreadsheets with a working online marketplace. Neither had written a line of code.

Both built on Emergent, a platform where a person describes the software they want in plain English and the system designs it, writes it, and runs it. About 70% of its users have never coded. Its roughly 200,000 paying customers spend about $600 a year each, for work that once meant hiring a development shop, and investors valued the company at $1.5 billion this week.

Earlier tools of this kind produced little more than simple websites; the newer ones aim at software a business depends on to operate. More than 12 million applications have been built on Emergent in the year since it launched, and the company says more than half of its paying customers made something now critical to running their business.

The skill once needed to run a ransomware attack has largely fallen away. Criminals now use everyday AI tools to write convincing phishing messages, pick the valuable files out of what they steal, and even negotiate with victims, work that once needed a trained crew.

Attacks in the first half of 2026 ran about 20% above a year earlier and have held near 2,500 a quarter, a level researchers now call the new normal rather than a spike. The AI here is not doing anything exotic, the security firm GuidePoint found; it lowers the cost of tasks people already did, so far more operators can run campaigns at once. Anthropic has separately reported the first intrusion campaign it saw run mostly by AI, with the human crew stepping in only at key moments.

The targets are rarely the giants: the Verizon 2026 Data Breach Investigations Report found 96% of ransomware victims were small and mid-sized businesses, the firms least able to absorb the hit.

On July 15, Wells Fargo gave its financial advisers, their support teams, and client associates a chat tool called AI Teammate. Ask it a question in plain language and it pulls answers on products, workflows, and processes from the bank's central adviser system, work that used to mean hunting across separate applications or asking a colleague.

Wells Fargo's headcount has meanwhile fallen for 24 consecutive quarters, to 197,000 people, down 79,000 over six years, even as revenue and client assets kept climbing; assets in its wealth arm rose 15% in the quarter. Chief executive Charlie Scharf credited efficiency from technology and AI, and said the same investments are "improving productivity" and lifting adviser hiring and retention.

The bank is putting AI closest to its most human, relationship-driven work, and still expects to employ fewer people overall. Its finance chief told investors Wells Fargo "should be able to run" with fewer staff than it has today.

Why do so many AI projects stall after the tools are bought? Increasingly the answer is people. The engineers who can take a capable model and build it into how a business actually runs are scarce, and the model makers are now competing directly for them.

This week Anthropic, with Blackstone, Goldman Sachs, and other investors, launched Ode, an AI services firm valued at about $1.5 billion that sends small teams of experienced engineers inside client companies to build and operate their systems. It follows OpenAI's own version, called The Deployment Company. Both aim at the buyers the large consultancies court: mid-sized banks, regional health systems, and manufacturers that want AI but employ no one who can deploy it.

Ode's leaders say demand for these teams far outstrips supply. For an operator, that reorders the cost of an AI project: choosing the tool is the easy part, and the people who can install it are now being bid for by Anthropic, OpenAI, Deloitte, and Accenture at once.