When it comes to artificial intelligence, the federal government and industry are both past the hype cycle.
It might not feel that way, but AI is well beyond the usual eight- to 12-month rise and fall of any new technology, said Max Claps, a research director for IDC Insights.
“All of the vendors and the federal government executives that I talk to are very much getting a grasp with reality, which for the federal government buyers means narrowing down the number of use cases that they can or want to scale and put in production,” Claps said during Federal News Network’s AI and Data Exchange.
In the December 2024 update to the AI.gov website, the Office of Management and Budget detailed more than 1,700 AI use cases across the federal government. Claps said the number is less important than what the agencies said they are doing with the capabilities and how they are gaining a return on investment and beginning to scale AI.
“I’m very intrigued by two things. The first one is there’s a little bit of a shift across the swim lanes. When I took a look at the 700 use cases a year and a half ago, they were a lot around data operations. There were a lot about IT and software development. Some, of course, had to do with business functions, but a lot of them were internally facing — a lot of human resources chatbots and so on,” he said. “Now, there’s many more mission use cases. We’re still probably in the crawl stage for many agencies, but it’s interesting to see how it’s drifting down into where it can really make an impact for the public.”
Federal AI uses cases come in a handful of types
Claps said most of the government’s AI use cases fall into one of five what he called swim lanes:
- Automation of business functions: HR, procurement and finance areas
- Automation of business tasks: mission specific applications
- Data operations: data preparation, data ingestion, data tagging and data pipelining to create the basis for automating tasks and enabling business users, analysts, policy experts to build insights and make decisions
- IT operations: service desk, provisioning of computing, observability and cybersecurity, and threat intelligent tasks
- Software development and deployment lifecycle: code documentation, legacy code refactoring and business requirements documentation
Claps said that as he analyzed agency AI uses cases, more moved from the pilot to operational stage.
“If you have an agency that has a lot of retired use cases, you may say they failed miserably with AI when actually that’s a positive indication. Potentially, it means they have a good AI innovation process in place that says this is a gate and we have 20% to 30% of our AI use cases that did not pass that gate because they did not prove their value so let’s retire them,” he said.
“If there’s an agency that only has operational use cases but no retired use cases, it means either they stopped innovating and they’re only trying to sweat their existing assets or they have too much on their plate and they’re trying to operate too many use cases. They’ve not thought of reprioritizing and saying, ‘Maybe this one is not delivering, and let’s repurpose the resources onto something else that may deliver more value.’ ”
Industry helps agencies adopt mission-driven AI
On the other side of the coin is the vendor community’s hype cycle path.
Claps said although many federal contractors initially wanted to push generative AI as the “best thing since sliced bread,” now they are more focused on government-specific use cases. He said this is an example of why AI has moved through and beyond the hype cycle in the federal sector.
“They are starting to talk about government-specific agents and consumption models, instead of expecting the government to buy an extra storage deck upgrade from them. They embed genAI capabilities into the next release of their standard products because that is a much better way for the government to consume their tools,” Claps said.
“Of course, the product companies are working with professional services firms to help the government realize ROI and scale the use cases that really matter. We’re really past the hype cycle, and everybody wants to talk about the practical implications.”
Given this, Claps said the next big thing will be the wide adoption of AI agents to support users in a variety of ways.
He shared the example of an agent that can capture insights from conversations during virtual meetings and then provide insights to inform the participants’ decision-making.
“There are, of course, agents being embedded in case management systems, agents that can draft emails and notifications for beneficiaries of certain services and programs,” he said. “I’m looking forward to the stage where agents are going to be an orchestration of different algorithms that can really close the loop around complex processes like child welfare, child protection types of processes or other services.”
Discover more articles and videos now on AI and Data Exchange event page.
Copyright
© 2025 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.
