AI In Local Governments: Global Expenditure To Reach USD 78 Billion By 2033

The World Governments Summit has issued a study arguing that artificial intelligence in local government could reshape how cities run daily services. The report says adoption is still limited worldwide, despite proven tools already improving licensing, traffic control, waste management, and public safety in several major urban centres.

The document, titled ‘Adoption of AI in Local Governments: Harnessing AI for Strategic and Operational Excellence in Local Governments,’ is produced with Arthur D. Little. It sets out a broad framework for AI-enabled municipalities, using global examples to show how data-led systems can be integrated into routine operations and long-term planning.

Financial projections in the report highlight the growing scale of AI in the public sector. Global spending on AI-based government solutions reached US$12.6 billion in 2023 and is expected to hit US$78 billion by 2033. The Government AI Readiness Index now covers 193 national administrations, signalling stronger expectations on public bodies to redesign services using AI.

Despite this spending outlook, actual deployment in city administrations remains modest. The report cites Bloomberg research showing that just 2% of surveyed local authorities have achieved full implementation of AI tools. Most municipalities are either testing small pilots or planning projects, rather than running AI at scale across core services.

{TABLE_1}

The study highlights several operational examples already in place. Abu Dhabi uses the virtual assistant TAMM to automatically process business licences and reduce manual checks. In China, Hangzhou relies on Alibaba’s "City Brain" system to direct urban operations, including traffic flows. Accra in Ghana deploys AI tools to identify and monitor illegal dumping locations.

Further case studies span the USA, UK, Australia, Brazil, and South Africa, showing how AI can target complex city issues. Pretoria has adopted AI-powered digital twins to optimise waste collection routes and schedules. São Paulo’s "Smart Sampa" platform supports public safety and policing, using facial recognition technology to assist law enforcement and strengthen security monitoring.

Barriers and solutions for AI in local governments

The report notes that many local governments face high entry barriers before they can copy these examples. Key obstacles include limited budgets, data privacy concerns, gaps in regulation, complex operational structures, and shortages of specialised digital skills. These pressures make it difficult for city leaders to justify large-scale AI deployment, even where potential benefits are clear.

To respond, ‘Adoption of AI in Local Governments’ proposes action across five main areas. It calls for strategic partnerships and regional cost-sharing to ease funding pressure. It stresses strong data governance, transparent communication, and privacy safeguards to build community trust. It recommends early dialogue with regulators, clearer policy frameworks, investment in staff skills, innovation-focused work cultures, and flexible systems that adjust quickly based on user feedback.

Dr. Raymond Khoury, Partner and Public Sector Lead at Arthur D. Little Middle East, stated: "Embracing AI at the local level could be transformational for cities worldwide. The obstacles to implementation are significant, but real-world use cases demonstrate what is possible when strategy, vision, and investment align. The task facing local governments is formidable, but they do not need to tackle it alone. Shared efforts can go a long way in building the cities of tomorrow."

The World Governments Summit report positions these recommendations as a practical guide for municipal leaders, including those in the Middle East. By combining partnerships, robust governance, skilled teams, and responsive systems, the study suggests that local authorities can gradually expand AI use, moving from small pilots toward wider adoption that supports more efficient and resilient city services.

With inputs from WAM