Companies are rethinking AI use as usage-based pricing raises costs

In short: Businesses are getting bigger and less predictable AI bills as AI vendors move from flat fees to pay-per-use pricing.

What's going on

Many companies spent the past few years encouraging staff to try AI tools freely, with few limits on cost. That is now changing, according to a Financial Times report. Since the start of 2026, large AI providers have been pushing customers away from flat monthly fees and toward usage-based pricing, which is more like paying for electricity by the unit.

At the same time, more companies are trying “AI agents”, which are AI tools that can do longer, more complex jobs on their own instead of just answering a question. These systems tend to use far more “tokens” (small chunks of text or data an AI model processes, like paying per word read and written). The result is that costs can jump quickly.

Uber said earlier this year it had spent its full 2026 AI budget by April. The company responded by limiting employees to $1,500 per month for an AI coding tool. In a KPMG survey of 2,145 business leaders in May, nearly half said they had scaled back use of AI agents because costs outweighed the benefits.

Some companies are adding internal controls. Atlassian has put caps on how many tokens each employee can use within a set period, with managers approving requests for more.

What to watch

More companies may try to avoid “vendor lock-in”, which is when switching providers becomes hard after building work around one company’s tools. Some are also looking at open-source AI models, which can be free to use and run on a company’s own computers, although they may take more effort to set up and maintain.

Source: Financial Times