States’ cautious AI strategies will become ‘harder to justify’ as tech improves, new report claims

The public is skeptical about the role that artificial intelligence should play in providing government services, and efforts by technology officials to deploy AI have so far been “cautious and uneven,” according to a report published Thursday by the National Association of State Chief Information Officers. But that caution, the report warns, will become harder to justify as AI’s capabilities improve.

The report, which draws on the opinions of 48 state CIOs, 300 other government employees and 1,000 U.S. residents, concludes with recommendations for the purportedly inevitable move to AI that state governments will make. (And, as the report’s findings show, states have begun to integrate AI into a growing number of their services and functions.) The “imperatives” for state IT officials, as NASCIO calls them in the report, include strengthening governance, designing systems that prioritize users’ needs and “starting with strategy and sponsorship.”

“Leaders at the highest levels—especially governors and agency heads—can remove barriers, accelerate alignment and reinforce urgency,” the report reads. “Without this mandate, [generative AI] efforts risk fragmentation and stagnation.”

Nearly half of residents said they still prefer human interactions over digital alternatives, and 63% said they prefer receiving in-person services. “Most consumers don’t trust AI when it comes to government services,” the report notes, adding that many said they are willing to forego the convenience afforded by digital technology. Meanwhile, about three-quarters of government employees said they’ve noticed that the public expects digital services to improve in quality.

And many state technology officials view AI as a way to improve their services — 90% told NASCIO they believe generative AI will “enhance citizen experience.” State governments have been cautious in adopting generative AI over the last several years, drafting detailed policies that keep in mind the possibility of introducing harmful new biases or privacy violations into their processes. But they’ve been slowly integrating AI into low-risk functions all the while and have increasingly turned to AI to keep pace in the cyber arms race.

The report shows that while CIOs are considering a broad range of uses for AI, the most common uses to have already been deployed include virtual agents and other chatbots, “frequently asked question” documents and language translation service for websites. AI uses that are either in early testing phases or are being considered for future use include citizen alerts and crisis communications, surveys and feedback analysis, and forms and applications.

Only 6% of CIOs said their states have fully “scaled” AI to their agencies, while 50% said they are currently “developing” AI for wider use. Thirty-one percent said AI is in the “defined” stage, meaning they have a strategy and maybe some low level of deployment. And 13% said their states are in the “exploratory” phase.

State CIOs won’t be able to deploy AI all by themselves, of course, and the state government employees who will help do much of the implementation are more optimistic than not about AI. Sixty-two percent said they believed AI would reduce their workloads by automating routine tasks, and a majority responded affirmatively both when asked if they believed AI would improve their work experience overall and if they believed AI would improve customer service.

Though AI deployment may be uneven across states, CIOs seem to be largely in agreement on the best strategies for overcoming their top challenges. Ninety-two percent of CIOs named taking a “phased approach” to AI deployment as a useful strategy to overcome their challenges, such as the public’s wavering trust and workforce shortages. Ninety percent said that improving data quality will help, followed by 79% who noted the importance of policies and governance.