As well as impacting many different industries, AI has the potential to be used in various areas of the public sector. In the UK, for example, 18% of all jobs are in the public sector, and the proportion is large in other countries too – 16% in the USA, 20% in France, 29% in Sweden. Japan has an AI-driven earthquake prediction system that analyses seismic data in real-time and can issue alerts, finding a 70% improvement in accuracy. The notoriously bad traffic in Sao Paolo in Brazil has been given an AI hand in the form of a smart traffic system that adjusts traffic signals and suggests alternate routes, resulting in a 25% reduction in travel time. Canada used machine learning to analyse financial transactions for suspicious activity and identify patterns of tax evasion, yielding millions in additional and formerly unpaid taxes.
There are many such initiatives happening around the world. Chatbots can help citizens navigate the public sector bureaucracy, something that has been successfully tried in Singapore. A UK government chatbot helps businesses understand regulations. In Seoul the city uses smart bins, powered by solar power, to compact bin contents, allowing bins to be emptied less often. This has resulted in an 83% reduction in waste collection costs, and has virtually eliminated the problem of waste bins overflowing. More controversially, the EU has employed AI systems, at least in a pilot project, as border guards, asking questions and analysing facial gestures to see whether travellers appeared to be lying, in which case they were referred for further checks.
Governments have many of the same issues as private sector companies, though without the profit motive. They want to be efficient, save money and improve services to their citizens. Just as with private companies, they also face challenges in the form of dubious data quality, limitations in data governance and a lack of skills in AI. They have additional issues in the form of a need for transparency, a (usually) greater emphasis on fairness and equality, and eliminating bias in systems. The same issues that affect the introduction of LLM technology into companies apply to the public sector: having to deal with the thorny issue of hallucinations, having to integrate with legacy systems, a lack of workforce skills, and the issue of public trust.
Governments need to plan carefully when considering rolling out AI projects. The UK government has defined ten principles to be used for such projects, including ensuring that staff know how to use AI securely and lawfully, understand AI limitations and that the right tool is used for the right job. Such a set of principles seems to be a good starting point. Just jumping on the AI bandwagon to grab headlines is likely to end badly. In September 2025, the Albanian government announced an AI procurement minister. Given the dismal job that AI agents can make of running even a minuscule business, and given their record in negotiating, this seems highly experimental and is likely destined for failure. Widespread staff training in AI literacy for public sector workers will be an important basis for success, as is the careful monitoring of AI projects, including learning from mistakes and sharing best practices. The private sector has enough difficulty in rolling out AI projects, with an MIT 2025 study showing a dismal 5% success rate for AI projects. Government projects tend to have additional constraints compared to private sector ones, and so are likely to encounter all the same issues, but more besides, since there is an understandable additional emphasis in the public sector on fairness and eliminating bias in systems. There are definitely opportunities to use AI in public sector projects, as we have seen from some of the success stories highlighted. The key is in matching the opportunity to the most suitable form of AI technology, and then ensuring that each project carefully considers the factors that will help it succeed, such as data quality, AI literacy of staff, quality assurance, testing and careful monitoring of the project. If governments act thoughtfully and responsibly, then they should be able to use AI to help them bring greater efficiency into citizen-centric public services.
