Kenneth Glad
Companies are weighing new tools against familiar business pressures. The next phase is about control, not novelty.
The AI rush has moved from technology teams to finance meetings. After a year of rapid rollouts, companies are asking a harder question: What does heavy use actually deliver?
In Denmark, similar concerns are surfacing among companies using enterprise AI tools.
Frederikke Saabye, director for digitalisation, technology and telecoms at Dansk Erhverv, the Danish Chamber of Commerce, told TV 2 Denmark that some firms bought wide-ranging licenses before setting a clear business case.
“People experience time savings, but very few can isolate the effect in the business figures,” Saabye said.
Global firms are pulling back
The concern is not confined to Denmark. According to TechCrunch, Uber had already exhausted its 2026 AI coding budget by April, creating an uncomfortable gap between spending and measurable business gains.
The issue was not simply that employees were using AI. It was that the higher costs were difficult to connect to stronger financial results, making the business case harder to defend internally.
“It becomes harder to defend because it is not free. AI is not free,” Uber chief operating officer Andrew MacDonald said, according to TV 2 Denmark.
Walmart has taken a more controlled route. Bloomberg, via Yahoo Finance, reported that the retailer capped staff use of Code Puppy, its in-house AI tool, after demand rose sharply among employees.
The tool had been introduced to help with tasks such as spreadsheets and presentations, but heavy usage meant the company had to place limits on tokens, the units used to measure AI computing activity.
Usage can become the goal
Amazon’s experience shows how incentives can distort adoption when companies reward usage itself. According to Financial Times, the company removed an internal AI leaderboard after some workers increased activity to improve their rankings, adding avoidable computing costs.
The ranking system had been intended to encourage employees to experiment with AI tools. Instead, it reportedly pushed some staff toward higher token consumption, even when the extra activity did not appear to improve the work.
“Please don’t use AI just for the sake of using AI,” Amazon senior vice president Dave Treadwell told the staff, according to Financial Times.
GitHub has also adjusted its model as AI coding tools become more demanding. In a company post, they said Copilot is moving to usage-based billing because long agentic coding sessions can require far more computing power than brief assistant queries.
The shift means customers will be charged more closely according to actual token use, rather than only through a flat request-based structure. For developers and companies, that makes AI spending more visible, but also less predictable.
Results remain uneven
Chris Reed, senior director of IT finance at Priceline, described the pressure to TechCrunch: “It’s like the crack-cocaine epidemic.”
“They let you try it to get you hooked on it, and now you’re kind of beholden to it,” he said.
OpenAI CEO Sam Altman has acknowledged the shift. “I think we can find ways to help people. It went from, at the beginning of the year, being a problem that was never mentioned, to suddenly being a big problem,” he said, according to TV 2 Denmark.
A Bain & Company study found that only 7 percent of surveyed companies have fully autonomous AI agents in production.
For employers, the message is blunt: Adoption numbers are no longer enough. The tools now have to justify the invoice.
Sources: TV 2 Denmark, TechCrunch, Yahoo Finance, Financial Times, GitHub, Bain & Company Study
