Inside GSA’s AI Transformation: How a Chatbot Sparked a Cultural Shift

Original broadcast 6/11/25

Presented by Carahsoft

Artificial intelligence doesn’t always begin with grand ambitions—it often starts with solving a simple, persistent problem. At the General Services Administration (GSA), that problem was routine administrative tasks eating up valuable time across the workforce. What began as a pilot project for a chatbot to support employees has become a foundation for GSA’s broader AI journey. Zach Whitman, the agency’s Chief AI Officer and Chief Data Scientist, shared insights at the AI for Government Summit on how GSA’s pragmatic, people-centered approach is paying off.

For Whitman, the key was starting small and learning fast. GSA spent nearly a year testing public generative AI tools on non-sensitive tasks, observing how employees interacted with the technology. What they found was striking: a well-designed chatbot could help with 70–80% of common low-value tasks, like generating bullet points, summarizing content, and drafting documents. GSA’s leadership saw the potential to save minutes per task—but more importantly, hours per week—across thousands of employees.

That insight led to the development and deployment of a chatbot made available agency-wide. Now in active use by roughly 30% of GSA’s workforce, the tool is already reshaping how employees approach their daily work. Whitman emphasized that adoption wasn’t driven by mandates or hype. Instead, the agency focused on enabling workers to discover the tool on their own terms and measure how it affected productivity. "We didn’t expect the volume of organic demand we’re seeing," Whitman explained, "especially the appetite for deeper integrations like API access."

Indeed, what started as a simple interface has evolved into a launchpad for more complex AI applications. Developers across GSA are now using APIs to connect the chatbot to their coding environments and robotic process automation (RPA) pipelines. These integrations allow AI to support more specialized workstreams—like contextualizing data, optimizing forms, or interacting with legacy systems. As employees become more fluent with the technology, they’re also becoming more creative in applying it to their domains.

To support this growth, GSA has built a robust framework for monitoring AI tool usage and evaluating model performance. Whitman underscored that success isn’t just about availability—it’s about reliability. His team assesses how different foundational models perform across areas like HR, procurement, and internal services. Accuracy, responsiveness, and safety are all tracked, with constraints added where necessary to protect sensitive information. That allows GSA to deploy models confidently in production environments, without compromising on mission integrity.

Equally important is GSA’s commitment to feedback. The chatbot itself includes thumbs-up/thumbs-down tools and open text fields for users to share their thoughts. In addition, Whitman’s team maintains a dedicated feedback email channel and frequently hears from employees about what’s working and what could be improved. This two-way communication helps identify opportunities to expand model capabilities or adjust guardrails based on real-world needs.

One of the most intriguing takeaways from GSA’s experience is how quickly the workforce is evolving in tandem with the technology. Whitman said his team was surprised by the growing sophistication of employee expectations. Staff began asking not just for basic assistance, but for agentic capabilities—where AI could string together multiple steps in a process automatically. That interest is now driving GSA’s exploration of multi-step workflows and task chaining, which could eventually serve as precursors to true autonomous agents.

GSA’s history with robotic process automation also played a role. Years before generative AI took center stage, the agency was experimenting with automation tools to reduce manual work. That experience helped shape an organizational mindset that welcomed AI—not as a replacement, but as a partner in getting work done more efficiently. Employees already understood that technology could offload repetitive tasks, and AI is now deepening that value by adding reasoning and content generation.

Looking ahead, Whitman envisions a future where GSA connects its AI tools directly to agency-specific data stores, allowing for highly contextual responses. That next phase will require careful model evaluation, privacy controls, and even tighter integration with GSA’s digital infrastructure. But with a solid foundation and a responsive workforce, the agency is well-positioned to lead by example.

The GSA experience demonstrates that successful AI adoption isn’t driven by novelty—it’s driven by mission needs and user trust. By embedding AI where the work already happens and measuring what matters, GSA has created a model for responsible, scalable, and impactful innovation across government.

Key Takeaways:

• GSA’s chatbot adoption grew organically, now used by ~30% of the workforce.

• Feedback and evaluation frameworks ensure accuracy, safety, and performance.

• Employee demand is driving deeper integration and exploration of agentic AI.

This interview was featured in the program Innovation in Government recorded on location at the Carahsoft AI for Government Summit. To watch the full program, click here.