As agentic artificial intelligence (AI) gains traction across the public and private sectors, one line of questioning emerged repeatedly at MeriTalk’s Mission Room today: “Can it? Should it? May it?” For federal agencies experimenting with systems capable of taking action, officials say the focus should be less on the novelty of the technology and more on how – and when – it should be governed.
For Jennifer Franks, acting chief technology officer and director of the Center for Enhanced Cybersecurity at the U.S. Government Accountability Office, those questions capture the shift from AI as an information tool to AI as an actor in agency workflows.
“The last few years we’ve been asking, what can AI tell us? But now we’re going to have to ask, what can that AI do?” she said.
For federal agencies, answering that question means looking beyond technical capability to determine:
- Can it? Before automating a process, agencies need to determine whether the organization and its underlying systems are ready
- Should it? Technical capability does not mean an AI system should be allowed to act independently. According to Franks, an agent’s autonomy should correspond to the consequences of its actions
- May it? The final question focuses on whether an AI agent has the authority and security controls necessary to act. Franks said agencies should think of AI through a cybersecurity lens – as an identity capable of exercising authority
The three-question framework does not envision humans remaining in the loop merely to approve every action an AI system takes. Instead, Franks described a future in which technology handles more routine, high-volume work while people concentrate on the parts of government work that require judgment.
“We can let the technology handle the problem … but we can let the people handle the ambiguity, the ethics, the professional skepticism, the relationships, the exceptions,” Franks said.
That makes workforce development as important as technology deployment.
“We want to be thinking about not replacing, but how do we retain our talent and how do we motivate them to stay within this emerging tech space,” Franks said.
Franks said the agencies best positioned for the future will not necessarily be those deploying the most AI agents. .
“They’re going to be the organizations that understand where their agents belong and what they do not,” she said.
Throughout conversations at MeriTalk’s Mission Room, federal and private-sector leaders similarly emphasized that the rise of agentic AI will require strong human leadership alongside technological capability.
Across discussions, officials pointed to the need for leaders to define where AI can add value, establish clear boundaries for autonomous action, and ensure people remain accountable for the decisions and outcomes that matter most.