GSA Weighs Cost Model for USAi Platform in 2027

The General Services Administration (GSA) is working with federal agencies to determine how to charge them for its USAi platform beginning in fiscal year 2027, as officials seek to understand the platform’s value and long-term support requirements.

GSA launched its USAi platform last year at no cost to agencies, allowing them to experiment with artificial intelligence (AI) capabilities without having to maintain duplicative platforms.

According to GSA’s budget document for fiscal year 2027, GSA is planning to start charging agencies for USAi as part of a “cost-recoverable model.”

“Right now, we’re working with the agencies to understand what would a cost model for them look like,” Zach Whitman, chief data scientist and chief AI officer at GSA, told MeriTalk on the sidelines of the 2026 ITVMO Annual Summit on May 6. “Because what we want to do is we want to provide a solution that they can see being a useful addendum to their current portfolio.”

Whitman explained that USAi is not designed to be an “end-all, be-all” or compete directly with industry. Instead, he said, the platform provides agencies with a baseline level of access to foundation models.

“A lot of agencies will have one or two hyperscaler availabilities, which limits the number of models that they can potentially reach,” Whitman said. “So, the value proposition for us is just making sure that they have availability to the leading labs’ models in a way where they can look at our evals, they can ensure that it’s safe, they can ensure that it’s fit for their purpose, and then also keep an eye on all the telemetry that’s occurring on that system.”

Whitman said the ongoing USAi trial has been critical in helping GSA determine whether agencies are seeing meaningful returns from the platform.

“We want to understand what the value of this thing is to them and make sure that we’re able to support that value with a staffed program office that would not produce a sub-quality product,” he explained.

“It’s really about understanding what ROI they’re seeing on the platform, and then how big our team needs to be to support them,” Whitman added.

During a panel discussion at the summit, Whitman said one of the central challenges facing agencies and vendors alike is determining how to price AI access when usage patterns vary widely.

“The OneGov deal and USAi both are trying to attack this inertia problem, of ‘I don’t want to buy everyone a license to just try this thing out.’ There’s an unknown quantity of, what is the ROI of giving everyone a chatbot?” Whitman said.

“There are attempts to try to figure out what the value proposition is for that, but fundamentally, we don’t know,” he said, adding, “So, … in the USAi deal, the concept was, ‘Well, let’s just try to give everyone access to this thing, so we can get some data to understand and quantify what the value proposition is.’”

Whitman said usage data collected so far shows a wide gap between heavy and light users across the federal workforce, which reinforces the need for agencies to move away from flat-rate licensing approaches for AI tools.

“We need to make sure that we’re paying for consumption,” he stressed. “And incredibly important is you need to control your API consumption at a central point, because that stuff can get out of hand real quick, depending on what you’re doing with it. So, we’re heavily focused on this.”