In March of this year, the National Association of State Chief Information Officers (NASCIO) published a forward-looking report on the rise of agentic AI in state government. Notably, AI has surged to the top of the association’s annual top-ten list of policy and technology priorities, displacing cybersecurity from its 12-year reign as the number one priority.
The report lays out a five-phase evolution, with NASCIO finding that most states have reported activity in the first phase of “assistive generative AI.” From there, the framework advances to “context-aware generative AI” followed by “task-level automation.” Agentic AI arrives in phase four with “stateful, multi-step workflows,” in which it “manages progress across time, steps, and systems.” In the final phase, “adaptive and proactive agents” are implemented.
As expected, state IT leaders are cautious about the deployment of various AI tools. The NASCIO report highlights some early pilot and sandbox efforts with the Commonwealth of Virginia’s plan to use agentic AI to reduce regulatory burden as the most ambitious example. In January, the Commonwealth’s Office of Regulatory Management (ORM) published a progress report on this initiative.
In that report, Virginia declares that it launched the “first-in-the-nation agentic AI-empowered regulatory reduction pilot,” putting itself “out ahead of essentially every other state in the AI space.” As a result of that pilot, agencies “streamlined 35.70% of regulatory requirements” and “cut over 12 million words from their guidance documents, representing a 49.69% reduction.” However, the report does not detail how specifically agentic AI workflows were used to achieve these results.
A review of GovSpend’s PO data shows an emergent $200 million niche market specifically for agentic AI offerings. Market researchers should look more broadly, as significant SLED spending on a range of AI tools can be found in licences purchased from companies such as Google, Microsoft, and Anthropic. Agentic development functionality is generally embedded in advanced subscriptions to cloud services, productivity suites, and other types of operational software. These offerings represent billions of dollars in sales annually.
One of the best ways for vendors to track SLED agencies’ progress through the NASCIO phases of AI evolution is through tools such as GovSpend’s Meetings module, where indicative discussions can be discovered. A particularly instructive discussion was held at the end of April by the Placer County (CA) Board of Supervisors, which heard a report from the county CIO on how “departments are beginning to move beyond basic AI tools, such as content generation or simple co-pilots and are now exploring systems that can take action on their behalf by automating workflows or supporting decision-making for staff.”
With these insights, vendors can keep tabs on this highly fragmented and fluid market for AI solutions.
About the Author: Chris Dixon
