The untapped potential of AI in the public sector

How county technology leaders can tap into the benefits of AI in local government

Artificial intelligence (AI) is continuously in the news—with new capabilities (and new risks) introduced almost every day. Among government agencies, many technology leaders are hesitant to embrace its usage; meanwhile, employees are already using it without true oversight. Despite the risks, a mindful approach to using AI in local government, and in particular generative AI, can provide many benefits to your agency.

While the terms are often used together, AI and generative AI offer different benefits. AI is the automation of tasks through computer systems, where little to no human intervention and oversight is required. Examples include speech recognition, language translation, and security cameras. Generative AI is the development and deployment of artificial intelligence systems that can generate novel content by relying on large language models, data sets, and other methods of machine learning (ML), including inputs from humans. Examples include ChatGPT, Microsoft’s Copilot, and Google’s Bard (now Gemini).¹

As an IT leader in local government, whether county, city, or regional, it is critical to understand how AI can be evaluated and used in your environment. Although it is hard to find the time and capacity to address the use of AI in the public sector, getting started will allow you to move from “firefighting” mode to “forest management” mode. With the right technical and policy steps in consideration, you can start building your framework for AI usage by public agency employees and residents.

Here are the 4 actionable things IT leaders can do to maximize the potential of AI in local government:

Step 1: Identify where AI exists in your agency today.

Step 2: Research and prioritize the data sets needed for top AI use cases.

Step 3: Decide what AI policies or guidelines you will update or create for employees around AI usage.

Step 4: Plan your roadmap for technical support of AI usage.

Identify where AI exists in your agency today

Before you can fully utilize AI in the public sector, you must start by running diagnostics to identify current AI usage on your network. Network traffic can identify which users are going to sites such as ChatGPT. Tools like ChatGPT are open use, which means anyone can create an account with their agency email address and use it. Users are frequently unaware that all of the prompts and data generated within the tool become public domain, and might violate records retention and public disclosure policies for their agency. The volume of data might seem small, but it’s important to understand what data is being placed in these large language models (LLMs).

Your security team will also have tools to determine where activity is happening. If your agency uses Microsoft 365 applications, ask your administrator to check the settings for Copilot usage. Be aware that if the plan is to simply turn it off, the settings return with every upgrade. Considering Microsoft pushes updates monthly (or earlier if there is a security update), this becomes an additional burden on limited IT staff.

If your IT team is small, ask for support from your vendors. Many of them have added AI diagnostics and functionality to their products, and would be happy to support a usage audit. There is federal support as well via the Center for Internet Security’s Multi-state Information Sharing and Analysis Center (MS-ISAC) and other agencies; services may be available gratis, or with a minimal fee.²
Research and prioritize the data sets needed for top AI use cases

Understanding the health of the data sets needed to train AI models for key use cases is critical for a successful deployment. When considering how to implement AI in local government, and which data sets to use for AI training models, it can be overwhelming to consider all of the types of data, their location, and their accessibility. Pick 1-2 areas to start that align with top Freedom of Information Act (FOIA) requests, key processes within top services (property assessments, vital records, and courts data come to mind), or high priority initiatives from elected officials that require extensive data and analytics.

Artificial intelligence can resolve many integration issues, but it won’t change the need for accurate data cleansing and inventory first. Set aside time for data reviews as part of tracking technical debt, and partner with key champions in the agency to promote the value of good data.

Anyone who is invested in informatics, records retention, or public disclosure will be a great ally in influencing other leaders. Consider automatic deletion processes for key records retention policies, rather than relying on staff who do not consider data management a key part of their role.
Decide what policies or guidelines you will create or update for employees around AI usage

Once you know where to find AI in local government, how it’s currently used by employees, and how you would like it to be used, take the time to update any relevant agency policies or guidelines. Some counties will find a separate policy is needed for the use of generative AI in local government, others will want to update existing policies with the appropriate, up-to-date language. It is critical that you don’t assume current policies, such as the IT Acceptable Use Policy, will cover AI use cases.

There are considerations for privacy, security, data governance, compliance, records retention, and public disclosure requirements, such as the FOIA. The federal government is updating their AI policies and guidelines as AI evolves; states, counties, and cities also must stay aware of the intersection of governance needed to craft effective frameworks.

Leverage policy guidelines from various local government associations, such as the National Association of Counties (NACo), your state association of counties, or county IT leaders. There may also be state policies that should inform your position. Remember, this doesn’t have to be a 45-page document. Start with the expectation that employees will leverage AI responsibly and to be mindful of the data they are using, in order to protect the county. Inviting a set of key business leaders to co-create a policy can go a long way when it comes to increasing buy in and reducing risk.
Plan your roadmap for technical support of AI usage

Whether your county plans to fully or partially embrace AI in local government, IT staff will need new awareness and skills to support its use. Take advantage of free, online training for the products already in your IT environment.

Encourage staff to explore the capabilities of existing applications, both as an end user and as a prompt engineer. Decide if you will allow employees to use outside LLMs, such as ChatGPT, or if the expectation is for them to stick to applications within the existing tech stack, such as Microsoft 365 products.

Larger agencies may benefit from the creation or expansion of their engineering team to include experts in ML and robotic process automation (RPA). Other agencies might outsource development with an established vendor or native AI startup. Whatever your approach, a roadmap is needed to address usage for critical business processes.

Many times, the more mundane administrative processes will net the biggest return in employee efficiency and productivity. Encourage your IT staff to think creatively about these types of processes, in partnership with their business counterparts. Start with a small set, then plan for extension into all priority business areas for the county.