GSA Positions Itself as a Federal AI ‘Enabler,’ CAIO Says

The General Services Administration (GSA) sees itself as an enabler for other federal agencies’ AI use as it works to align with new presidential directives, the agency’s Chief Data Scientist and Chief AI Officer Zach Whitman said Tuesday at GovCIO Media & Research’s AI FedLab in Reston, Virginia.

“We see ourselves in GSA as an enabler for the federal workforce, for the federal complex generally,” Whitman said. “And we want to have the best-in-class tools for the staff, no matter what.”

Streamlining AI Procurement

Whitman said that while the presidential transition shifted AI priorities to promote speed, several core themes across administrations remained the same: ensuring safety, transparency and the ethical use of AI. GSA, he said, is an “enabler” for streamlining AI procurement for other agencies to help meet agencies’ shorter planning horizons.

“We want to make sure that we are doing right by the other agencies in terms of looking at procurement, consolidating it easier and faster for more scalable acquisition,” Whitman said. “We want to focus on ways in which we can empower other agencies to buy better, buy smarter and buy more efficiently.”

Whitman added that GSA is looking to reduce the time-to-market for accessing industry-leading AI tools. This necessitates a flexible approach, moving away from vendor lock-in to allow agencies to choose the best AI model for their specific mission.

“Locking in with a single vendor just doesn’t comport with that line of thinking. [Agencies] want to have flexibility because of how dynamic this space is. It’s really important that we have the ability to switch between the best in class model for the best use case,” Whitman said. “The variety of the use cases that we’re seeing are huge. The ultimate goal for us is really about making sure that we can have access to the latest and the greatest wherever possible.”

Enabling AI Solutions to Scale

GSA envisions itself as a benchmark, Whitman said, for effective AI acquisition and deployment, and the agency plans to make its processes and platforms available to others.

GSA is working on a scalable platform accommodating diverse AI models and providers. The rollout of internal AI chatbots has revealed a significant appetite for such tools and a subsequent “rapid maturation of the employee base” as they identify new ways to integrate AI into their workflows.

“One initiative that we’ve been doing … is making sure that we have a scalable platform that allows for a variety of different AI models and providers to have a seat at that table,” Whitman said. “There’s a huge appetite for that.”

The next step, according to Whitman, is to foster “AI native workflows,” where processes are designed with AI tools in mind from the outset for agencies throughout government.

“We have to serve, in part, as a benchmark for what good acquisition looks like, what good deployment looks like and make what we do available to others,” he added.

Fostering a ‘One Government’ Ethos

Collaboration is central to GSA’s AI strategy, including sharing models. Whitman pointed to the Department of Energy labs’ development of a custom large language model for scientists as an inspiring example of federal innovation. He said that GSA plays an “infrastructural role” in facilitating the sharing of code, models and compute resources across agencies.

“How can we, as a ‘one government’ ethos, really lean into that aspect of our scale, lean into the aspect of our talent and maybe empower others from different agencies to use that LLM?” Whitman posited.

AI progress and adoption hinges on the engagement of willing industry partners, Whitman said. GSA aims to create a “level playing field,” where various foundation model providers can offer their services, fostering a dynamic marketplace and lowering the barrier of entry for smaller organizations, he noted.

“We need to figure out how to at least crack a window on and begin to open those back up,” said Whitman. “This space is very early [in development] and there’s a lot of new players in the game that may not be thinking about selling to government,” said Whitman.

Prioritizing AI Practicality

Whitman noted a shift towards “grounded practicality” for AI uses compared to the previous year. Early exposure to public AI tools for non-sensitive tasks has fostered familiarity and a move towards more programmatic integration of AI across government.

“People are starting to think about that next step outside of those transactional interfaces … and thinking about it more programmatically,” he said.

The agency’s GSAi is a platform concept encompassing a chatbot, an API platform and an administration console for monitoring safety and evaluating model performance. Whitman expressed excitement about the platform’s focus on quality metrics tailored to specific agency functions, like procurement. This approach aims to move beyond basic safety concerns to ensure AI tools provide accurate and relevant advice.

“It focuses on the quality aspect. Instead of focusing on only safety concerns, we are asking ‘are we introducing biases?’ Is it teaching me how to disown my kids by providing bad legal advice?’” said Whitman. “We are focusing on the quality of the performance of each model based on the specific function areas of the agency.”

Boosting Data Quality Across Government

GSA is highlighting the importance of data quality and metadata in its pursuit of AI adoption. Whitman said that AI is a catalyst to improve data practices throughout government.

“We want to use AI as a kind of Trojan horse to backdoor [better] boring metadata and data practices … There’s no better way to demonstrate the quality – or the lack thereof – of your data stores by putting an AI engine in front of it and letting it hit its head against the wall,” said Whitman. “We have ignited a lot of conversations across the organization about like, what can we do to improve our data stores.”

Whitman said that AI adoption is inevitable for government and GSA is leading the way for its staff and all of government.

“We believe it’s a market norm. At this point, every workplace is going to have some sort of capability,” he said. “We need to make sure that we’re doing right by our employees.”