How Do Chief AI Officers Help State Governments Innovate?

What Do Government Chief AI Officers Do?

As AI becomes more prevalent, there are a small handful of states that have appointed officials “who can oversee the development of policies, bring together stakeholders and advise agencies,” says Amy Glasscock, NASCIO's program director for innovation and emerging issues.

“To have someone doing this full-time means they have the time to really learn about AI and stay abreast of the seemingly constant changes and developments in a way that others may not be able to do, given the workload in state government in general,” she says.

Josiah Raiche, director of the division of artificial intelligence for Vermont’s Agency of Digital Services, has been in his position for a little over a year following the 2022 passage of legislation that made Vermont the first state to have such a role.

“The reason we decided to do that was because we saw that the ethics and policymaking side of it was really big, but we also wanted to have someone who was responsible for adoption,” he says.

READ MORE: Contact center AI boosts citizen services for states.

Raiche says states can have officials who do both, or two officials who split the duties of managing regulation and adoption, but that states do need to have leaders focused on these issues. These officials brief lawmakers, CIOs and other key decision-makers on AI policy and use.

In November 2023, Vermont issued guidance for how state agencies can use tools such as OpenAI’s ChatGPT and Google’s Bard (now Gemini), and on Jan. 25, 2024, the state released an updated inventory of all automated decision-making systems in state government and how they are used.

Nikhil Deshpande, Georgia’s chief digital and AI officer (the AI role was formalized in September 2023), says he sees his job has having two prongs, similar to Raiche.

One prong is managing AI governance frameworks and developing and enforcing standards to “provide some guardrails around the use of AI technology.” The second prong directly flows from that, and it is to answer the question of whether the right AI tools are being used to solve the right problems.

“A lot of times, I have seen excitement around solutions but very little effort to understand and solve the right problems,” Deshpande says.

How Has the Position of Chief AI Officer Evolved?

The role of chief AI officer is still so new that there hasn’t been much time for it to evolve. Yet it’s on the minds of state CIOs, Glasscock says, with many debating whether to appoint such an official.

“Some think it’s where their state will go in the future, and others think that it doesn’t make sense to hire someone that specializes in just one type of technology,” she says. “So, while I think we’ll see more, I don’t think it’s a given, like every state having a CISO. Only half of states even have a chief privacy officer at this point.”

Deshpande says the evolution of these kinds of roles all boil down to “having a product vision.” Machine learning models were being used in government before ChatGPT. Now, the job is focused on determining how to effectively implement AI solutions in government. That will require a major focus on data quality, he says, and how well AI tools perform using that data.

AI tools, especially generative AI chatbots, rely on having large amounts of high-quality data fed to them so they can produce high-quality responses to queries. “If you don’t feed it good data, then it is not going to be as effective,” he says. “And that’s really the crux of that role. It’s to make sure that when we implement it, we implement it such that we focus on how our data is structured, how our data is sanitized. And then as we start training the model, understanding how well it performs and if it really addresses the problem that we’re trying to solve.”

Deshpande says he sees the CAIO role as being extension of chief digital officers’ roles, with a big focus on ensuring a good user experience and that AI tools are built with users in mind.

DISCOVER: The roles of CIO vs. CTO vs. CDO in government.

What Advantages Does a CAIO Provide State and Local Governments?

Chief AI officers deliver significant benefits for state governments that want to expand their use of AI tools — not least of which is taking the pressure off CIOs and other leaders to be responsible for every AI-related decision.

“While so many people are going to be involved in setting AI policies and overseeing AI in the executive branch — the CIO, CISO and chief privacy officer — it does take some of the pressure off of these people to be the AI expert at the state, given how many other duties they have on a daily basis,” Glasscock says.

In the early stages of building out an AI program, a CAIO brings an executive mindset to setting a vision for AI use, Raiche says. That official can talk to CIOs, CTOs, lawmakers, the state attorney general and other officials to craft the best approach.

CAIOs can also be very valuable in demystifying AI tools for officials who might not be as familiar with them and addressing any concerns those officials might have, Raiche says. That can lead to productive conversations about the desired benefits and guardrails for AI solutions. 

Until state governments sort out issues related to the ethics of AI use in state government — for example, how it might be used to determine how benefits are awarded — and until the technology is mature enough, states are going to need CAIOs, Deshpande says, to ensure AI tools are designed well and are compliant with state cybersecurity and ethics policies.

“But also,” he says, “the actual implementation piece of it is very important, because once things are offered, when things are out there, the genies are out of the bottle, and then it might be too late to reverse course on some of those things. I think, especially for government, it is very important that we do the right things, and we do it the right way.”

Are There Examples of CAIO Successes in Government?

It’s still early days for CAIOs in state government, but both Raiche and Deshpande point to some early successes.

In Vermont, in the wake of catastrophic flooding in July 2023, Raiche’s team was able to quickly launch a generative AI system that basically functioned as a search engine that cities and towns could use to hunt for state contracts. That allowed municipal officials to quickly search for and find emergency state contracts they could tap to buy gravel, for example.

Because the contracts were created in haste, some might have had small typos or might not have been categorized properly, Raiche notes. A traditional search engine wouldn’t have worked well with messy data. However, using OpenAI’s models, Raiche’s team quickly trained a tool using the new contract data to create a brand new chatbot.

“It’s not very shiny, but we were able to build that,” he says. “I think it took us two days. We were able to get the job done because we had already been doing a bunch of work around this space. We had already been doing pilots for building chatbots to build intent detection systems, all kinds of stuff like that.”

LEARN MORE: How state CIOs expect AI to benefit digital services.

In Georgia, Deshpande has created an enterprise AI program within the Georgia Technology Authority focused on establishing an AI governance framework for state agencies, creating an innovation lab to serve as a central hub for AI research and development, and developing an AI education program for state workers to ensure they are AI literate.

Georgia is working on an AI ethics playbook to ensure that “the transparency and accountability aspects of all the AI applications are very clear, and agencies have a way of disclosing that,” he says.

Georgia hopes to have the basic physical version of the innovation lab to whiteboard solutions and design new AI tools. The state is also going to have a virtual component of the lab for agencies to test AI tools, Deshpande says.

“We are encouraging agencies to look at their processes, to look at what some of the success factors of service delivery are that they envision, and truly see where we still have opportunities to enhance certain services,” he says.