Local governments have become one of the biggest users of AI without much notice. Not the flashy kind, chatbots, planning tools, and back-office systems that answer resident questions, process permits, and flag broken roads.
The scale is already large. A city in Texas uses a virtual assistant to handle about 16,000 resident enquiries a month, which helped it avoid spending 1.8 million dollars to expand its call center. A council in Sweden put image recognition on its garbage trucks to spot potholes as they drive past. A city in China uses AI to monitor plants and generate maintenance orders when pests show up.
The economics behind this are hard to ignore. One estimate for councils in England and Wales found that AI could take over more than a quarter of everyday local government tasks, worth roughly 8 billion pounds a year in staff time if adopted everywhere. That is the kind of number that gets a budget committee's attention.
But speed has a cost. A recent benchmark test asked several popular AI models more than 22,000 questions that mirror what a real resident would ask a council website, things like how to apply for a benefit or which tax deadline applies to a small charity. The models gave wrong answers on basic eligibility rules, incorrect legal requirements, and grants that do not exist in certain regions. A resident who trusted that advice could lose money or miss a deadline through no fault of their own.
This is the part that should worry any business that deals with local government, whether as a supplier, a taxpayer, or a resident. A council does not get to blame the AI when it gets something wrong. The legal responsibility stays with the government, the same as if a human employee had made the mistake. That means every AI tool a council buys is really a liability the council is choosing to take on, whether it has the governance in place to manage it or not.
A study spanning five countries, Australia, the United States, Spain, Hong Kong, and Saudi Arabia, found a consistent pattern: adoption of AI is running ahead of the rules needed to govern it. The researchers point to six gaps that keep showing up: unclear accountability, staff who do not understand what the AI can and cannot do, weak data protections, low transparency about when AI is being used, little input from residents, and no plan to retrain workers whose jobs are changing.
For businesses that sell technology or services into the public sector, this is an opening. Councils that get governance right, clear rules on when a human has to review a decision, plain disclosure to residents that they are talking to a machine, will move faster and face fewer scandals than councils that bolt on AI without asking who checks it. Right now, most are still figuring that part out.
The tools are real and they work. The question every council, and every supplier selling to one, still has to answer is who checks the machine's work before it becomes a resident's problem.