From Strategy to Execution: How Agencies Are Making AI Work

Technology leaders from the National Telecommunications and Information Administration (NTIA), NASA and the former Defense Digital Service (DDS) talked about the gap between ambitious AI strategies and the practical realities of implementation at GovCIO Media & Research’s AI Summit Friday in Tysons, Virginia.

Alan Rosner, acting CIO at NTIA, is tasked with carrying out major elements of the federal AI Action Plan.

“The biggest challenge that we face is bringing AI expertise into our workforce and also adopting an AI culture,” he said.

But Rosner said they are finding opportunities for both employees and other government agencies that interact with NTIA’s systems to get the most out of AI capabilities including incorporating chatbots to answer questions or check the status of a project.

“So as opposed to knowing the fine details of all of our business processes, and who do you have to talk to, avoid calling somebody inside the bureau, those types of things, we want to make the access to information as easy as possible to the various government agencies, that interact with our systems to get our data,” he said.

NASA Workplace and Collaboration Services Director Cara Rose said the agency uses Copilot, an AI-powered assistant from Microsoft, to improve efficiency and safety for the Orion team.

“There’s a lot that goes into sending four … astronauts to the moon,” she said. “Safety’s the number one concern. There’s so many checklists and flight readiness reviews and artifacts that have to be completed, and it seems like not that big of a deal, but putting Copilot into the hands of the Orion project team so that they could complete all of their flight readiness reviews in half the time that it took them before, Copilot is really improving efficiency while maintaining the safety of our astronauts.”

NASA is now exploring similar AI-enabled efficiencies in service desk operations, where chatbots could resolve routine tickets automatically, freeing staff for more complex work, Rose said.

Jennifer Hay, former director of the now-defunct Defense Digital Service, encouraged organizations interested in putting AI into action, to make sure they have a good data foundation in place. She said organizations often request an AI solution before fully understanding the problem and recommends that they build a process to collect “clean, good data” in an automated way to get them on the path to AI.

“We really drove that focus on the outcome. What is the decision you are trying to make? What decision is AI going to help you come to? And then we backed it up to put in place automated processes to get them to that point where they could eventually come in and bring in, an AI solution that would really, truly speed up their process,” she said.

As far as what’s coming next for AI, Red Hat CTO John Dvorak said open source and open standards are the way to innovate around AI. For government, open source offers a practical advantage: it’s far easier to share capabilities across agencies and with international partners when tools are built on open standards.

“If we all build on an open standard, then it’s much easier to work not just within government, but across NATO, across other organizations, outside the United States where we want to innovate together and iterate together,” he said.