Not long ago, companies were practically begging employees to use AI more. Some built internal leaderboards to reward heavy usage. Others warned staff that skipping AI tools could hurt their careers. The message was clear: use it or lose out.
Now the bill has arrived. And it turns out that getting thousands of employees to run AI tools all day on tasks like converting PDFs into PowerPoint slides adds up fast. According to TechCrunch, companies are now doing the opposite of what they encouraged just months ago: rationing AI usage and trying to stop workers from burning through token budgets on trivial work.
The shift is real, and it is happening quickly. What started as a race to adopt AI has turned into a scramble to control costs before leadership starts demanding answers.
Accenture is one of the clearest examples of this reversal. 404 Media obtained leaked audio from an internal meeting where Accenture’s agentic AI strategy lead, Justice Kwak, told staff the company was hitting a wall. “We’re hitting this inflection point where AI is becoming material to the cost structure,” Kwak said in the recording. “Spend is becoming very unpredictable; and leadership, especially at the CFO, COO, and CIO level, are still asking the question of whether they’re getting value from what we’re spending on in the context of AI.”
That last part is the crux of it. CFOs are asking whether the spend is worth it, and right now, many companies cannot give a clean answer.
The irony at Accenture is hard to miss. The firm reportedly warned employees earlier this year that they would “risk losing out on promotions” if they did not use AI tools. Now the same company is trying to stop those same employees from using AI for low-value tasks that eat through token reserves without producing meaningful output.
This is not just an Accenture problem. Across the industry, a pattern is emerging:
- Companies pushed hard for AI adoption, often without clear guidelines on what counted as good use
- Employees responded by using AI for anything and everything, including tasks where it added little value
- Token costs piled up faster than expected
- Finance and operations leaders started questioning the return on that spend
- Now the same companies are pulling back and trying to build guardrails they should have had from the start
The broader market is also feeling this. What some are calling an “AI selloff” has hit several AI-dependent businesses hard in recent days, with memory chip makers taking a particular beating. Investors are starting to ask the same questions that CFOs inside companies are asking: where is the actual return?
For years, the AI industry ran on excitement and potential. That worked when the technology was new and the costs were abstract. But tokens are not abstract. They show up on invoices, and those invoices are now landing on desks where people want to see results, not just adoption metrics.
The companies that come out ahead will be the ones that figure out quickly which AI use cases actually move the needle and which ones are just expensive busywork. That is a harder problem than it sounds, especially when the culture inside many organizations was built around using AI as much as possible rather than using it well.
