Data Analytics

NASCIO 2024 Midyear: Speedy GenAI Adoption Sparks Data Quality Concerns

State IT officials see challenges in data management, job classifications and hiring.

Mickey McCarter is the senior editor of StateTech Magazine.

The speed at which generative artificial intelligence is being adopted has raised concerns about good data management and data quality, said Alaska CIO Bill Smith Tuesday during the midyear conference of the National Association of State Chief Information Officers.

Speaking on a NASCIO conference panel in National Harbor, Md., Smith said that generative AI has tremendous potential to improve citizen engagement and overcome government workforce limitations. But states must first ensure that they have quality data to empower the technology.

Smith compared the urgent need to improve data management to the necessity of enhancing cybersecurity for mass remote work during the pandemic.

“It caused us to spend a lot of time on basic cyber hygiene. We should have been doing that all along, but it really focused our attention,” Smith said. “The same thing is happening to data. We are not talking about anything novel when it comes to data, but we are seeing an increased focus on data hygiene.”

Should states begin to examine generative AI use cases, they may realize that they do not have good-quality data, and they will lose time “fixing the data,” Smith said.

“The urgency around AI may cause us to say, we cannot fix the underlying data, so let’s create a bespoke, isolated, curated data set for this AI use case. That is equally terrifying for me because then downstream, you must maintain 100 different pockets of data.

“Hopefully, this inspires us to get the resources to do some global hygiene and really operationalize quality data management moving forward so that we have sustainable, quality data for our AI systems,” he said.

Click the banner below to see how digital transformation can benefit state governments.

For Data Demands, About Half of States Are Stuck in Reactive Mode

During the NASCIO panel, Chris Estes, EY’s U.S. finance, operations and technology leader for state, local and education, shared some preliminary results from a NASCIO and EY report on data management, which he anticipates will be released in about a month.

Forty-six states responded to the survey, and 96 percent of those surveyed said that the increased adoption of generative AI is driving a need for better data management.

North Carolina Chief Data Officer Christie Burris said the survey results show that states must “proceed with caution.”

“With the speed we are seeing in GenAI , if we fail due to the quality of data, we are going to fail very quickly,” Burris said.

The survey asked respondents how they would rate the maturity of data quality in their organizations. Six states indicated they were “aware,” the lowest level of maturity, and 26 said they were “reactive.” Twelve said they were “proactive,” and two said they were at a level where data quality was “managed.”

“Reactive means you have a use case or a set of questions you are trying to answer with data, and you see additional questions you must ask later. That’s reactive mode. We find ourselves in that all the time,” Burris elaborated. Proactive, on the other hand, means that officials are looking ahead with an eye on data quality and anticipating questions before they arise.

The maturity of data quality management differs widely between programs in Alaska, Smith said.

“In Alaska, it’s a little bit all over the map. We have some areas where we maintain data within the state, and it is very proactive,” he said. These areas generally have significant compliance requirements. “And then we have other areas, where nobody thinks of it until we need it for a project. It just sits there until someone says we need this for this purpose.”

“I classify us as reactive,” Smith said of the state’s data management overall. “By and large, we don’t have a good, formulaic way to make this an operational norm. But I think the attention around AI is going to help.”

State Governments Have Ill-Defined Data Job Classifications

The NASCIO-EY survey asked states what types of data roles are prevalent in their IT organizations. Here is how some of the 46 states responded:

Database administrator, 35
Data analyst, 23
Data architect, 19
Data engineer, 18
Data scientist, 16
Data steward, 10
Data literacy manager, 7

“I think we are going to see more deliberate attention being paid to the skill sets involved at all stages of data management,” Smith said about the results. “In Alaska, our database managers do it all. They are expected to own it, soup to nuts; it’s data, it’s yours.”

Alaska soon will adopt new IT job classifications and descriptions. “We are getting ready to move with a more modern job classifications for our IT workforce starting in the new fiscal year. It’s taken us about four or five years to build that. We don’t have a lot of data scientist references because it’s a very fast-moving train,” Smith said.

Burris agreed that generative AI is a pressing need for delineating roles in data management. “We don’t have job descriptions to recruit the right talent,” she said, and so the state relies on vendor partners to fill the gaps.

“In the near term, we are going to have to fill these holes with partners,” Smith said.

Keep this page bookmarked for our coverage of the NASCIO 2024 Midyear conference. Follow us on X, formerly known as Twitter, at @StateTech and the official conference Twitter account, @NASCIO. Join the conversation using the hashtag #NASCIO24.

pabradyphoto/Getty Images