Chief Data Officers Talk Data Governance in Preparation for AI

It is becoming abundantly clear that artificial intelligence is poised to usher in the next major step in agency modernization. With state legislators and industry experts encouraging agencies to embrace AI and maximize their efficiency, the state will likely see a major increase in its usage in the coming fiscal year.

During a panel on data governance at the Texas Association of State Systems for Computing and Communications (TASSCC) State of the State Conference on Friday, Department of Family and Protective Services (DFPS) Chief Data Officer (CDO) Amberle Carter and University of Texas at Austin (UT Austin) CDO Shiva Jaganathan advised agencies and vendors on how best to manage and protect their data, a step that industry experts stress is paramount before AI implementation.

The primary topic of Carter and Jaganathan’s conversation was data governance, which describes the internal data policies and standards of an organization. According to Jaganathan, good data governance is the key to good data.

“Any strategic play and metrics associated with those metrics can only be as good as the data,” said Jaganathan. “You have to have great data governance established so your data is curated and validated and all your decisions are based on factual accuracy. If you don't have good data governance, whatever strategic plan you're trying to achieve is going to be built on not-so-good data.”

Carter explained that one best practice that organizations should take to improve their data governance is establishing a single source of truth regarding industry terms and their definitions. According to Carter, doing so allows your data to be more accessible and usable.

“A lot of the struggle in terms of the use of data is that lack of an agreed upon business term,” Carter explained. “Because different departments would say, ‘Well it’s something that fits these certain criteria,’ but another division would say it was different criteria and say, ‘I don't really understand why your numbers aren’t the same.’”

Carter and Jaganathan also stressed the importance of identifying security challenges and minimizing risks, especially while using AI. For Carter, the human element of data governance is one to be considered.

“AI technologies such as ChatGPT, they do present challenges in that you can tell people, ‘Don’t put this employee’s record with their Social Security number in here and come up with something.’ They’re still going to do it,” said Carter. “So you're going to have to figure out how to monitor that and place appropriate guardrails and controls around the use of AI so we are responsible in our agencies. Are we following best practices to maximize our output and minimize the risks to the agency and the rest of the people that we serve to protect?”

Jaganathan agreed, pointing to role-based access as one safeguard that all agencies should be implementing and maintaining. Role-based access limits an individual’s access to the data they need to perform their role. According to Jaganathan, the most important step an organization can take to secure sensitive data upon a team member’s exit is immediately shut down their access.

“That is the most important thing,” said Jaganathan. “You’re talking about privacy; you’re talking about information that is not falling into the wrong peoples’ hands. And those are all achieved by role-based access.”