The next wave of federal AI adoption will come from government employees building solutions tailored to their own work, not from top-down IT mandates, according to Alexis Bonnell, head of AI adoption and deployment at OpenAI for Government. Speaking Tuesday at MeriTalk's Shift Happens event in Washington, D.C., Bonnell said the "built by user" approach gives employees the ability to create AI tools that match their expertise and daily tasks.
"For the first time, people were actually able to craft a relationship with knowledge on their terms, in their way," Bonnell said. "They didn't have to wait for me to look at requirements, and I think this is why built by user is so important."
How a "built by user" approach shrinks IT backlogs
Bonnell drew on her experience as chief information officer at the Air Force Research Laboratory, where employees once submitted 186 requests for different technology solutions. After she introduced a basic generative AI tool and trained users to work with it, the list shrank to just 36 requests. Employees built solutions around the information most relevant to their own jobs, eliminating the need for many custom IT projects.
Rather than treating AI as another enterprise system, Bonnell said agencies should view it as an extension of employees' expertise. "There is a reason you are in the seat you're in," she told the audience. "Thinking about AI as a way to add and to unleash the total potential of you is a much more successful way for not only leaders, but for each of us to understand how AI should be and exist in our space."
RAG before agents
Bonnell urged federal agencies to focus first on helping employees organize and interact with their own knowledge before racing to deploy autonomous AI agents. "A lot of times, I think we skipped retrieval-augmented generation (RAG) and got really in love with agents," she said. RAG allows an employee to provide their "knowledge universe" to a large language model, optimizing its output based on the specific documents, data, and context that matter to their role.
Bonnell's advice comes as federal agencies increasingly invest in AI for Government, seeking to modernize workflows through tools that employees can shape themselves. She said leaders often overlook the most impactful innovations because they emerge from employees closest to the work - not from strategic planning sessions. "I will tell you that almost every amazing solution that I've seen has come from the place that the organizational leadership least expect," Bonnell said, citing finance and procurement staff who redesigned their own workflows with AI.
Why this matters for government employees
Bonnell's message puts the responsibility - and the opportunity - squarely on public servants. AI does not change the expectation that they come to work "intentional and accountable and responsible," she said. Instead, it gives them a faster way to exercise their knowledge and judgment. The shift from waiting on IT to building solutions themselves could reduce bottlenecks and let employees solve problems at the speed of their own understanding.
"I can't wait to see what version 2.0 of each of you looks like when you're empowered by knowledge at speed and scale," Bonnell said.
Your membership also unlocks:
