The OECD's “Governing with Artificial Intelligence” report analyzes 200 real-world use cases of AI in the public sector
Share
Artificial intelligence (AI) has emerged as one of the most transformative forces of the 21st century, becoming a key part of the digitalization of governments around the world. The use of AI by public administrations allows to automate internal processes and public services, personalize citizen care, improve decision-making and predictive capabilities, detect fraud and enrich the learning and quality of work of public employees. According to the Alan Turing Institute, AI could automate up to 84% of repetitive transactions in public services, saving the equivalent of 1.200 person-years of work annually. Despite this potential, the adoption of AI in the public sector still lags behind the private sector.
200 cases have been analyzed: 57% of the cases are automation
The OECD has published a new report "Governing with Artificial Intelligence" where it has analyzed more than 200 cases of AI use in 11 key government functions. The results show that AI is most present in the provision of public services, the administration of justice and citizen participation, while its use is more limited in policy evaluation, tax administration and civil service reform. This distribution is explained by the variety of possible uses in some areas (such as public services) and by regulatory or technical restrictions in others (such as tax management or administrative reform).
The most common use cases are based on classic rules-based approaches or machine learning techniques, while generative AI (such as large language models) is still rare. The main benefits identified are process automation, improved decision-making, and anomaly or fraud detection.
The path to AI is not easy: most initiatives do not get beyond the pilot phase
Implementation challenges are notable: lack of specialized talent, difficulties in accessing and sharing quality data, financial costs, regulatory rigidity and old technological infrastructure. Many initiatives remain in pilot or preliminary stages (60%) due to these barriers.eresIn addition, the lack of monitoring and evaluation mechanisms makes it difficult to measure impact and detect risks.
Using AI has risks. Not taking advantage of AI also has risks.
There is no such thing as risk-free adoption of AI. Biased algorithms can generate adverse outcomes; inappropriate use can violate fundamental rights; lack of transparència and public understanding can erode citizen trust; and overreliance can widen the digital divide and propagate systemic errors. There is also the risk of job displacement if AI replaces, rather than augments, the capabilities of public employees.
Not taking advantage of AI also carries risks: opportunities for improvement are missed and the gap with the private sector widens. Governments must embrace AI to respond to growing citizen demands and strengthen trust in institutions. Waiting until all the unknowns are resolved can relegate the public sector to a passive and costly role.
Towards a trustworthy, people-centered public AI
To harness the benefits of AI and mitigate its risks, the OECD recommends that governments deploy:
facilitators: governance, data, digital infrastructure, talent, investment, agile procurement processes and capacity for collaboration with non-governmental actors.
Safeguards: standards, guides, frameworks transparència and responsibility, and supervisory bodies.
Participation: mechanisms to involve citizens, civil society and businesses in the design and use of AI.
The OECD encourages prioritizing high-benefit, low-risk applications, and moving towards transparent, ethical, and safe AI. Most governments still lack processes to comprehensively measure the outcomes and impacts of AI, which should be a priority.
Reference: OECD (2025), Governing with Artificial Intelligence: The State of Play and Way Forward in Core Government Functions, OECD Publishing, Paris, https://doi.org/10.1787/795de142-en
