AI pricing models are evolving—and it might cost companies more

The nature of AI is sometimes you have to put a stop to unnecessary behavior—and cash (or lack thereof) can help make that happen.

GitHub is turning to metered pricing. The company, a subsidiary of Microsoft since 2018, explained the move in a blog post on April 27; its Copilot product will adopt a token usage model starting June 1.

It’s a move that should have been expected, said Harness Field CTO Adam Arellano. Old AI pricing models, including all-you-can-prompt for a set monthly price, were unsustainable, and that pain was going to be felt sooner or later. OpenAI already introduced token-based pricing for its popular ChatGPT model.

Economics of service are like a natural ecosystem, Arellano told IT Brew, likening it to population control in the wild.

“You get rid of all the mountain lions, the deer population is going to explode, and then it’s going to rubber-band back because there’s not enough food,” Arellano said. “What initially happened was these providers of agentic coding were getting so much cash that they didn’t have to fight for survival, they didn’t have to actually do anything useful or be lean or careful.”

Pattern shift. The company’s plan, which it’s calling GitHub AI Credits, is likely to change how organizations approach AI-assisted coding, Ensar Seker, CISO at SOCRadar, told IT Brew in an email.

“Until now, many teams treated these tools as ‘unlimited productivity boosters,’ but metered usage introduces governance, optimization, and ROI discussions into the software development life cycle,” Seker wrote. “Developers may start reserving AI assistance for higher-value tasks such as debugging complex logic, code review acceleration, infrastructure-as-code generation, and documentation rather than routine autocomplete.”

Once you start charging people for a product, there’s an expectation of service—and that’s something that companies using tokenized AI charging are going to have to reckon with. As Mike Wehrs, TieTechnology COO, put it, the rapidly evolving nature of AI means that usage-based payments could result in companies spending more money than they anticipated as they try to figure their way through problems and ambiguities.

That’s not going to provide you with a good bang for your buck, especially if you’re charged for the extra tokens and credits you need to spend to solve an issue.

“What good is 40 more tokens going to do when you couldn’t get it right with the first 100 tokens that I bought?” Wehrs asked rhetorically, adding, “That’s the kind of stuff that they don’t take into account, and they don’t have the customer support policies thought through to deal with that yet.”

Niche behaviors. Ultimately, Arellano told IT Brew, this pricing move may have come about somewhat haphazardly, but that’s no surprise: AI is moving fast, and many tech entrepreneurs have been more focused on creating products than the business aspects.

“They make cool things quickly and get them out to people fast, and then the realities of the economy and the economies of this come up,” Arellano said. “What I would say in their defense is that there is nothing that has ever moved like this or this quickly, except for crypto.”