To adopt AI, states scrambling for data governance, survey shows

State technology officials say that data quality is increasingly important, yet few states actually operate a data quality program, according to a preview of survey results the accounting firm Ernst & Young shared at a technology conference on Tuesday.

According to preliminary survey data of state technology officials presented by EY executive Chris Estes, 42 of 46 states ranked data quality as “important,” “very important,” or “critically important.” But 78% of states that responded to the survey said they don’t have a data quality program.

Alaska Chief Information Officer Bill Smith, who spoke during the session, guessed that the gap in data readiness is a result of states rushing to catch up with the demand for high quality datasets precipitated by the emergence of generative artificial intelligence.

“In the last 6-12 months, the priority and the importance [of data quality] rose and that’s not enough time to get a program going,“ Smith said. “It’s something we should have been doing all along.”

States are now rushing to establish AI governance policies, working groups and data frameworks that will allow them to use the emerging suite of generative AI tools increasingly pitched by the private sector and injected into existing software. Smith said his IT organization is among those not operating a formal data management program, but that some agencies in the state — like those that deal with medical data — operate formal data management programs as an artifact of their compliance work.

North Carolina Chief Data Officer Christie Burris, who also spoke during the session, said states will always have some low-quality data, but that the goal of data governance is to be prepared for new potential AI use cases as they emerge and prioritize the more important projects.

Burris referenced, but didn’t specifically name, a “high-profile” criminal case from the mid-2000s as an example in which the state was unable to find a perpetrator because the state’s data was distributed across many agencies, with no easy way to combine or share it. She pointed out later in an email to StateScoop that that case was the impetus for the North Carolina Government Data Analytics Center.

Trust, she said during the NASCIO conference, has been a key component of improving data governance inside state government, and today North Carolina operates data portfolios for health care, “child and family well-being” and a longitudinal data service.

“Data sharing moves at the speed of trust,” Burris said.

Ninety-six percent of state officials who responded to the EY survey agreed that increased adoption of AI and generative AI will “impact” the importance of data management. But states have a mixed ability to enact their data and AI programs: The survey showed that 26 states are “reactive” when it comes to data maturity, six states are merely “aware” of data, only 12 are “proactive” and only two states said their use of data is “managed.”

“We’re either going to have to push back really meaningful use cases for AI because the data is not ready,” Smith said. “Or we’ve got a really prioritized use case out there, in order to get past that hurdle, we create a bespoke set of data that is isolated and might be curated for that need. But then downstream I’ve got a hundred different pockets of data that are not necessarily correlated or not necessarily updated and it becomes a big burden.”

Burris said her data team is part of an informal AI and data workgroup that has recently demonstrated the importance of cooperating across agencies and various stakeholders surrounding state government.

“We can move their use cases through the framework and really think about implementation, but we’re planning to proceed with caution,” she said, adding that the first use cases in her state are “low risk” and don’t expose to the public to generative AI.

EY plans to release its finished survey results this summer.