GSA Develops AI Chatbot While Creating Frameworks for Government-Wide AI Acquisition
The General Services Administration (GSA) is taking a dual approach to artificial intelligence by developing its own AI chatbot, GSAi, while simultaneously creating frameworks for government-wide AI acquisition. According to Larry Allen, GSA’s associate administrator for government-wide policy and chief acquisition officer, “AI is already in the government, but it is going to get much, much bigger.”
GSAi is being used by the Federal Acquisition Service and Public Building Service to provide employees with quicker access to regulatory information and automate research tasks. Allen demonstrated the tool’s capabilities by explaining how it can process prompts such as understanding statutory requirements for government acquisition and identifying associated laws. This functionality can accelerate market research and aid in rewriting acquisition regulations. “Having to go through and research every little line item to see what statute it goes to. AI can assist with that,” Allen added.
The agency is also exploring how AI can enhance contract drafting and evaluation. “We’re also talking about how we can use it to make our contracts better, how we can put out better statements of work. GSA AI can help with that, and it can also help with evaluating some of the offers,” Allen said. While human judgment will remain central to GSA’s decision-making processes, AI is expected to play a supporting role.
In addition to its internal AI implementation, GSA is working on mechanisms for other agencies to procure AI technology. Two emerging models include subscription-based services and token-based models. The latter allows for more specific, on-demand access to AI capabilities. However, the novelty of the AI market poses challenges for procurement officials, particularly in determining fair pricing. “What is going to be key for contracting officers and others is how to identify price reasonableness,” Allen noted.
The AI market is still in its developmental phase across both commercial and public sectors. As the market matures, it is expected that identifying price reasonableness will become more straightforward. Allen drew parallels between AI adoption and the early days of computing, suggesting that successful implementations will drive wider acceptance. “No one in government likes to be the first to use most technologies… once the first two or three agencies tried it, and the cloud didn’t come down with thunder and lightning on it, it was okay for everybody else to use it too.”
While the government is not expected to mirror the commercial world’s AI adoption due to security requirements, Allen emphasized the potential benefits of AI. “The United States government has to have protections in place in order to ensure that its systems remain secure. But is there a tremendous amount of upside potential to make more use of AI? Absolutely.”
