Office workers are spending way too much on AI too

The AI push runs on setting venture capital money on fire and charging chatbot users way less than the bot costs to serve. The idea is that the customers will get hooked on the business value. Then the AI vendors can gouge them. And the vendors can try not to haemorrhage quite so much cash. So...

The AI push runs on setting venture capital money on fire and charging chatbot users way less than the bot costs to serve. The idea is that the customers will get hooked on the business value. Then the AI vendors can gouge them. And the vendors can try not to haemorrhage quite so much cash. So Microsoft, Anthropic, and OpenAI have all been moving their software customers off monthly subscriptions to token-based billing. Tokens have the added enterprise pricing advantage that the cost of anything is completely obscure. One source told Axios that one of their clients had spent $500 million in a month on Claude Code. We don’t know who it was — but the best guesses are Amazon or Uber, both of which are known to have run up stupendous bills. [Axios] But the customers are not so hooked they can just swallow their GitHub Copilot bill multiplying by a hundred. It’s not just the coders — it’s the ordinary office workers! They’re being told to AI up everything and give it that veneer of slop. That’s professional now. Unfortunately, it’s also expensive. Walmart said in early June that it was rationing its internal AI tool “Code Puppy”, which did office workslop as well as code. Code Puppy used to be unlimited. Now it’s not. [Bloomberg] Accenture has been having a rollercoaster ride. In February, Accenture was telling staff that getting a promotion would depend on hitting the chatbot. And if you didn’t, you were fired. This was really an excuse for layoffs — Accenture’s consulting business is badly down and its stock price has cratered over the past year. But Accenture employees responded to the incentive — and now the bill’s coming due. 404 Media got a leak of a meeting at Accenture: [404, archive] “We’re seeing from some of the data internally at least that it’s actually not our engineers that are driving the token consumption. It’s a lot of the non-engineers that are doing some of those behaviors […] you were talking about.” … Stuart Henderson … jokes he hopes Kwak didn’t just convert a PDF into images and then into markdown files. “I’m learning that’s one of the big token chewers,” Henderson says. “Turning PDFs into markdown: is that right?” If you order people to use a chatbot that doesn’t do anything useful, they’ll just point it at any old trash. Then the bill hits. The customer backlash is bad enough that Sam Altman at OpenAI is openly talking about price cuts. [WSJ, archive] But OpenAI can’t afford price cuts. They need revenue numbers to make their planned IPO look plausible. And Altman doesn’t have the sort of Elon magic that SpaceX ran their IPO on. Elon Musk isn’t convincing the markets either. The SpaceX stock price shot up — then back down a few days later. It’s been steady since, just below the offering price. The AI scam cannot possibly pay for itself from sales. AI turns out not to be critical for real business work. The use case for AI is doing things that should not be done. And there’s only so much market for that. Video — Podcast — Tradução para o português

Source: Pivot to AI — Published June 25, 2026 at 20:53 UTC — Category: Business