- Covington is launching a new AI initiative for city operations, the official policy for which was voted on this week
- The initiative aims to determine how AI could be used for city operations and possibly the public while also establishing clear guardrails and security polices
- The commissioners seemed open to the idea, especially in the face of city’s ongoing financial woes
The Covington Board of Commissioners voted this week to establish an in-house AI policy for city staff and city operations as one of the first steps in an institution-wide AI initiative dubbed Covington Operational Vision- Advance Innovation, or COV AI.
Essam Elgusain, the city’s IT manager, traced the effort back to a conversation he’d had with Mayor Ron Washington, who had asked Elgusain’s department to produce a draft AI policy.
After a few weeks, Elgusain said, his department was able to furnish one, but in that time, Washington had used an AI platform to generate a similar proposal.
“The proposal ended up being exactly the same,” Elgusain said Tuesday night. “The only difference was that it took him two days, instead of two weeks, to receive that proposal.”
This led to a bigger question, Elgusain said, “if AI can do this already, how can we as a city use it to work smarter, faster and better for our citizens with the limited resources that we have?”
What is usually referred to as generative AI, or artificial intelligence, is a cluster of technologies that uses complex mathematical formulas to generate (hence the name) a response based on a set of data inputs, usually referred to as training data.
Popular AI platforms like ChatGPT and Microsoft Copilot can use the entire internet as a training data set, or a user can upload an original data set from elsewhere to use as training data. Generative AI has several uses, including drafting and editing documents, data analysis and the automation of repetitive tasks.
Elgusian had recommended adopting an explicit AI policy at the caucus meeting last week. The policy will serve as a base for a broader, roughly year-long effort to train up the city’s workforce in AI platforms, identify areas in the city that would (and wouldn’t) benefit from AI use, establish oversight and governance of the platforms, and ensure the initiative remains both sustainable long term and working in the public interest.
The city’s department heads, as well as the mayor, have already been trained in AI proficiency, certifications that cost about $450 a pop. Elgusain put future costs for group trainings of lower level employees at about $5,000 each. He also said the city could potentially hire an AI consultant in the future to help with the rollout.
Although Elgusain and Washington seemed optimistic about the initiative, Elgusain admitted it carried risks. In fact, much of the policy adopted Tuesday night formalized guardrails against problems that afflict AI platforms themselves and against misuse by humans using them.
Even with the potential of AI, Elgusain said, “with that potential comes a set of responsibilities, ethical challenges, data privacy concerns and the need to make sure that the tools that we’ll eventually be using are transparent, and we’re using responsibly, especially at the local government level.”
The new policy establishes protocols for purchasing AI tools and sets up training requirements for the IT department. City workers would not be allowed to enter any sensitive information into an AI model, and the use of any AI product for work purposes would require sign-off from a department head. Employees would also have to fact-check any AI-generated outputs.
Other prohibitions laid out in the policy include the transmission of AI outputs containing city information in personal emails, the creation of deepfakes or other misleading imagery or media, the AI generation of any legal or contract documents “without human review and oversight,” the use of unlicensed products, the use of AI for hiring or disciplinary decision making and using copyrighted materials either as an AI input or in an AI output without legal permissions, among other strictures.
When Elgusain concluded his presentation, Commissioner Tim Downing asked him directly, “Did you use AI to help with your presentation?”
“A little bit, yeah,” Elgusain said.
Then Downing recommended restricting city AI usage to a single, controlled platform, rather than allowing multiple platforms (and potentially multiple payment agreements and the like) to co-mingle in city business.
“Having a singular entity like Copilot,” Downing said, “that we can put parameters and controls around tends to allow us greater limitation on the way that information is being used.”
Commissioner James Toebbe wanted to know if the city had used AI or even other programs to automate repetitive tasks. Elgusain said they did not use bots or pre-AI programs, but the department was using AI now to begin automating some tasks.
“I think with City Hall running short staffed, anything that increases employees’ capacity is great,” Toebbe said.
“As we continue on, I think it’s important for our workforce, as we’re moving forward, that they are able to identify some of the problems with AI, but I believe that it could be a savings for our community because certainly we can’t just keep trying to hire people,” Washington said. “We don’t have that kind of budget, but we should probably embrace technology as much as possible.”
You can read the full policy below.