The bills are coming in on AI tools, and suddenly human payrolls don’t look so bad.
Companies including Anthropic and OpenAI lured firms in with flat-rate fees and monthly subscriptions, but now they’re pivoting to pay-as-you-go plans, with usage calculated in units called tokens.
Must Read
Those tokens don’t add up to token amounts. We’re talking figures that rival salaries in the thousands, hundreds of thousands and even millions of dollars. And that money is being spent to design codes and applications to replace human work.
Arvind Jain is the CEO of Glean, a firm that provides businesses with AI and cloud computing support. He told CNBC (1) this situation is not only unexpected, but unprecedented.
“This is the first time ever that I can remember that technology costs the same as people,” he said.
In many cases, AI is not just costing the same as people, but more. Bryan Catanzaro, Nvidia’s vice president of applied deep learning, told Axios (2) that the “cost of computing is far beyond the costs of the employees” on his team.
Some firms are cutting their AI contracts and looking for cheaper alternatives. Others are bringing humans back.
Microsoft lets staff go — then lets Claude AI go
Last year, Microsoft announced it was prioritizing AI (3) investment — piling $80 billion in AI data centers and letting 15,000 people go (4) worldwide. This April, the company announced buyouts (5) in the U.S., targeting senior executives and those who might be nearing retirement.
Now they’re letting Anthropic’s Claude Code software go, too. Turns out Microsoft engineers blew through (6) the firm’s entire 2026 AI budget on Claude in the first few months of the year. The company is also doing more to reduce its reliance on OpenAI’s ChatGPT and other AI models.
It’s not that it’s doing away with AI. After all, Microsoft has built Large Language Models (LLMs) into every aspect of its business and promotes AI with Microsoft 365 Copilot. But the firm is doing its best to develop a proprietary AI software to avoid getting dinged on usage.
Prof G Markets co-host Ed Elson predicts (7) that a growing number of developers at firms and startups will opt for cheaper, open-source AI models than high-end Claude and ChatGPT, noting that made-in-China options are up to 30 times less expensive.
That could contain the soaring costs of developing AI applications at the front end. But Elson notes (7) that in some cases, the resulting AI products — the AI applications themselves — are costing more than the humans they were designed to replace.
That’s led to what observers call the “AI boomerang (8).”
Read More: BlackRock warns buying and holding the S&P 500 isn’t enough for retirement anymore — here’s why
The AI boomerang — bringing humans back … for now
According to an Orgvue (9) survey, 55% of employers who laid off workers and replaced them with AI regret their decision. Some companies are hiring humans back.
Turns out humans may be a better deal when it comes to building relationships, data judgment and quality control. One big example of that is customer service.
Gartner, a research advisory firm, expects (10) that 50% of businesses that laid off customer service reps or people in operation roles will end up hiring them back in the next year. Goldman Sachs research (11) in May found that human call-center agents could cost less than AI counterparts.
Klarna, the global fintech company, laid off 700 customer service reps (12) and replaced them with an AI chatbot. It looks like a great business decision at first, saving the company money and time. But customers were so upset with the chatbot that Klarna had to hire back the humans.
So we’ve reached a new chapter in the story of the AI revolution, one in which AI can cost more than people, and companies are looking for cheaper AI solutions, weighing whether to invest in AI tokens or drop more headcount — or even bring people back.
But it’s not the last chapter. “The ‘humans are cheaper’ era has an expiration date,” warns (13) technology analyst Rob Enderle.
“The cost of AI is likely to follow a trajectory similar to solar power or flat-screen TVs: an initial period of ‘only for the rich; followed by a precipitous drop,” he wrote in Technology News World.
He said some data suggests AI labor could be 90% cheaper in the next 10 years, particularly as short language models (alternatives to LLMs) increase and the infrastructure is built out.
“At that point, a human being’s time will become a luxury good — something people pay extra for, like hand-stitched leather or artisanal sourdough,” he wrote.
You May Also Like
Join 250,000+ readers and get Moneywise’s best stories and exclusive interviews first — clear insights curated and delivered weekly. Subscribe now.
Article Sources
We rely only on vetted sources and credible third-party reporting. For details, see our ethics and guidelines.
CNBC (1), (4), (5); Axios (2); BBC News (3); Cybernews (6); Profgmedia (7); The Interview Guys Blog (8); Orgvue (9); Gartner (10); Goldman Sachs (11); Medium (12); Tech News World (13)
This article originally appeared on Moneywise.com under the title: Humans suddenly look better than AI when it comes to company costs — at least for now. Here’s the surprising change
This article provides information only and should not be construed as advice. It is provided without warranty of any kind.