Whether they’re driven by Executive Order, shrinking budgets, or a desire to modernize the mission, federal agencies are turning to AI to improve service delivery to citizens and to their internal stakeholders. While nearly all agencies have a long list of AI use cases, many are waiting to operationalize them because they lack clear guidance on some very complex issues that could either make or break the success of their AI pilots.
To help guide agencies, Federal News Network’s Jason Miller gathered four experts from higher education, industry, and federal service to share their AI best practices.
Best Practice 1: Establish the Ground Rules
As with any new venture, establishing ground rules for application and deployment is an essential element for success. Jane Pinelis, Chief AI Engineer at Johns Hopkins University Applied Physics Lab, has used a combination of organizational culture and the principles of explainability, transparency, and traceability to incorporate AI into the organizations – which includes the Department of Defense – that she works with. In her conversation with Miller, she noted that “an organization’s culture plays a role in determining whether to use AI or not – or even where to apply specific capabilities.” Pinelis also noted that building comfort with senior leadership helps an organization embrace AI more quickly and effectively.
Best Practice 2: Turn Guardrails into Real World Gains
From her extensive work with federal agencies, MaryAnn Monroe, Vice President of the CX Accelerator at Maximus, has already seen how AI has improved service delivery and customer experience. From AI assistants that help agencies be responsive to citizen needs 24×7 to complex large language models that can help analyze petabytes of data to support those on the frontlines of service delivery, Monroe can share countless examples of how AI helps agencies modernize the mission.
One way that Monroe sees agencies deriving significant benefits from AI is in the creation and use of sentiment analysis. “Sentiment analysis lets agencies perform topic modeling to determine in real – or near real – time what subjects are most important to customers,” she explained. “Intelligence from sentiment analysis keeps an agency informed.”
Best Practice 3: Prepare for the Unknown
In a world that is changing so quickly, being prepared for the unknown is excellent advice. As Kathleen Featheringham, Vice President of the AI Accelerator at Maximus noted, “sometimes you can prepare in advance for customer needs, but what about the unexpected questions or issues that you didn’t anticipate? You need a way to capture that information and feed it back to agents when it comes up again.” In other words, AI can be used as a way to use data to get to better outcomes faster and build an understanding of trending topics to support human agents as they interact with citizens.
For Featheringham while most federal agencies are just getting started on their AI journeys, they need to be thinking about the next step – Agentic AI. Agentic AI is one of the essential tools for building next-gen and next-level service delivery models by allowing agencies to become even more responsive via automation. Agentic AI helps agencies breakdown workflows and then reintegrate them so they can deliver smarter, faster, more intelligent responses.
Best Practice 4: Field Test AI to Ensure it Delivers
Many good ideas have fallen apart when it comes to putting them into action. And undoubtedly this is a fear of any agency taking its first forays into AI. In the last part of the conversation with Jason Miller, Eric Hysen, the former Chief Information Office and Chief AI Officer at the Department of Homeland Security shared his experiences in making that step from theory to practice across the agency from work done with the United States Citizenship and Immigration Service to the Homeland Security Investigations organization, which forms part of the Immigration and Citizenship Enforcement bureau.
If you want to read and learn from his experiences, you can read it by downloading the Executive Briefing Insights here.