How to approach token cost uncertainty

Sponsored by

Header: 59% of professionals spend 11+ hours a week chasing information across systems.†

Copy: This guide breaks down how to connect context across tools and keep work moving.

†Quickbase, 2025

Like open-mic nights, it can sometimes seem like anything goes when it comes to AI companies pricing tokens.

Some organizations are pushing for more standardization around token costs. For example, as IT Brew has previously reported, the Tokenomics Foundation has been trying to “standardize” the value of AI.

Kirstie Tiernan, a principal and national AI lead at BDO, a national advisory firm, said that without a predictable, reliable cost structure for AI spend, finance teams and chief financial officers can become confused. This is potentially compounded by the sheer amount of AI vendors to choose from, as well as a lack of education around tokens and pricing.

“[Vendors] should be competing on their quality, on the accuracy, on the availability and the reliability at which they’re providing their services,” Tiernan said. “Without that level of standardization, it’s just so hard.”

Tiernan said that understanding the current environment around token pricing starts with understanding vendors’ price structures and monitoring your organization’s use of AI models.

From there, IT teams can educate their employees on tokens consumed and how that impacts the organization. An app that estimates token usage could help show token burn in near real time.

What pricing structures could look like. Meanwhile, organizations relying on AI just want predictable and stable pricing, Brandon Barbello, co-founder and COO of Archetype AI, told IT Brew.

Enterprises trying to understand what AI providers are selling them should ask about alternative types of token measurement, including tying it to value.

“My recurring advice to customers is that it’s important that they establish their own way of measuring success,” Barbello said.

For smaller IT departments that may have fewer resources for monitoring their AI spend, Tiernan added, professionals should consider vendors that offer observability tools, such as weekly spending reports.

“Once a month, that IT team is sitting down with finance, and you’re looking at the spending, you’re looking at those results together,” Tiernan said. “You have to take advantage of the tools they have, and the tools we have now for monitoring and usage and figuring out what is being used are way better than they were a few months ago.”

Barbello agreed, adding that evaluating and comparing vendors’ cost structures and monitoring is critical.

Subscription model. Babak Pahlavan, founder and CEO at NinjaTech AI, is against a pay-per-token approach, and that the industry should instead move towards subscription-based pricing models.

“Especially for work enterprises, you buying tokens is very unnatural,” Pahlavan said. “It’s an enemy of your margin.”

Tiernan said that, with fragmentation across the landscape, AI subscriptions would make budgeting a whole lot easier—but potentially cause additional concerns around AI model capacity, especially since subscription usage may be capped in some scenarios.

“As far as supporting the subscription…I think if you got really clear on the usage reporting and what happens when you need additional capacity, I think that could work,” Tiernan said. “Again, standardization is going to help because buyers are going to have to understand what that subscription gets them, as well.”