GSA is planning to bring its chatbot to the rest of government

The General Services Administration is looking into how it can implement its internal artificial intelligence chatbot across the federal government, the agency’s top AI and data official said Thursday, the latest indication that the Trump administration is planning on streamlining government access to AI. 

The new initiative marks the “next iteration” of the GSAi platform, Zach Whitman, the agency’s chief AI officer and data officer, said during a speech at the Digital Government Institute’s annual convention in Washington, D.C. 

The GSA rolled out GSAi internally in March after a lengthy research and development process, which involved an AI safety team that evaluated a number of major AI vendors. 

Like other AI chatbots available to the public, the tool was initially designed to respond to user prompts and assist in basic tasks. GSAi gives users access to a number of models, including ones from OpenAI, Anthropic and Google, and aims to increase workflow efficiency at the agency. 

This next chapter, Whitman said, is “one where other agencies could use what we have, use it in an isolated environment, use it for their specific purposes and own it in a tenant-based model.”

Whitman said GSA has spent “a lot of time working on our own problems” and considering “what would help us make buying decisions, which models work the best, categorizing the risks, safety concerns.” 

“Every single agency that we’ve talked to has the same exact problem set,” he continued. “We’re in active conversations right now with other agencies — how we can empower other agencies like ourselves to make better buying decisions.”

Should GSAi be offered to other federal agencies, GSA would still be the responsible agency for maintaining the infrastructure and analytics of the platform, Whitman said. 

A different federal AI official recently confirmed to FedScoop that integrating GSAi into other agencies’ AI tools was discussed at a recent meeting of the CAIO Council. The person said potential concerns are protecting individual agencies’ data and preserving security. 

At GSA, Whitman said the goal “is really just to make sure that agencies have an opportunity to try out these different models and make an informed decision based on our evaluation work and our practical experience.”  Some AI tools might not be the right fit for certain agencies, he noted, but they need access to the tools marketplace to make that determination. 

Whitman’s comments come just weeks after GSA Deputy Administrator Stephen Ehikian said that nearly half of agency employees are using GSAi every day. The GSA receives requests for new use cases daily, Ehikian added, hinting that more examples of government AI use will soon be released. 

Whitman told reporters Thursday that he hopes an expanded GSAi rollout will result in cost savings as it eliminates the need for some isolated agency AI contracts. 

“There’s so much duplicative work that’s going on,” he said. “Everyone’s building the exact same thing. Like we were just talking to another agency, the band, the architecture is the same. We all build the exact same thing, and it’s just a question of like, do we all need to do the same thing, or could we find some efficiencies?”

Whitman emphasized GSA is remaining “vendor agnostic” and exploring several major models, a pledge that comes on the heels of previous FedScoop reporting that GSA employees were testing Grok 3, the AI tool built by Elon Musk’s xAI. 

Days after FedScoop’s report, xAI announced Grok was available to purchase through GSA.

Whitman offered some new details about Grok on Thursday, noting that the tool is still in an evaluation phase at the GSA. 

“Because we’re thinking about this more from a measurement perspective, we’re not trying to make it — we’re not trying to put it into operation,” he said, adding that “the main focus is any major commercial model that’s coming up. We want to measure it to see how it performs, and we would make an adjudication for its use based on [a] safety team.” 

Agency members will evaluate Grok 3 and 4, and should the models pass specific testing, they will be evaluated by the agency’s AI Safety Team to ensure they are meeting agency standards, Whitman said.

Rebecca Heilweil contributed reporting.