Governments around the world are facing a difficult balancing act. They are expected to respond quickly to increasingly complex problems, from artificial intelligence and cybersecurity to housing shortages, healthcare pressures and climate resilience. Yet speed alone rarely produces good policy.
A growing body of research suggests that the quality of public policy depends less on how quickly governments act and more on how effectively they use evidence when making decisions. That conclusion sits at the heart of a paper published in OECD Digital Government Outlook 2026, which examines how governments are building the institutional capacity needed to govern artificial intelligence responsibly.
Rather than focusing on AI itself, the report explores something arguably more important: what makes governments capable of introducing new technologies without compromising accountability, transparency or public trust. As AI becomes increasingly embedded in public services, the findings offer a broader lesson about governance in the digital age.
The analysis draws on data collected from 36 OECD member countries, combining national policy responses, institutional surveys and comparative governance indicators. Instead of measuring technological capability alone, researchers examined the practical foundations required for effective public-sector AI, including governance structures, procurement practices, staff skills, regulatory oversight and mechanisms for measuring outcomes.
One pattern emerged quickly.
Most governments have made significant progress in developing AI strategies, with 35 of the 36 countries reporting that artificial intelligence is already being used in at least one area of government. Adoption has been strongest in administrative processes and public service delivery, where AI can improve efficiency and reduce routine workloads. More complex areas, such as policymaking and regulatory oversight, have seen slower uptake as governments grapple with higher standards of transparency, explainability and risk management.
The research also challenges a common assumption that governance slows innovation.
Across the countries studied, the biggest obstacle was not excessive regulation but uneven institutional readiness. While most governments had developed national AI strategies, far fewer had invested in the systems needed to implement them effectively. Data governance, procurement frameworks, digital infrastructure and specialised public-sector skills frequently lagged behind policy ambitions.
Perhaps the most striking finding concerned oversight.
Although every OECD country reported having at least one safeguard governing AI use, relatively few had introduced comprehensive operational controls. Only 39% required formal risk assessments before deploying AI systems, around one-third had internal review committees, and even fewer conducted routine post-deployment audits. Standards for algorithmic transparency were similarly inconsistent, despite increasing public expectations around accountability.
These findings point to an important distinction between having rules and having governance.
Policies can establish principles, but effective governance requires institutions capable of applying those principles consistently. That includes clear lines of accountability, ongoing monitoring, skilled public servants and the ability to evaluate whether policies are achieving their intended outcomes.
The same lesson extends beyond government.
Financial institutions, insurers, healthcare providers and large corporations increasingly face similar governance challenges as AI becomes integrated into lending decisions, fraud detection, customer service and operational planning. Boards are no longer asking whether AI should be adopted, but how it can be governed responsibly while maintaining public confidence and meeting regulatory expectations.
South Africa finds itself navigating many of these same questions. The country’s recently released draft National Artificial Intelligence Policy places strong emphasis on governance, institutional capacity, education and responsible innovation as AI adoption accelerates across both the public and private sectors. While the technologies themselves continue to evolve rapidly, the international evidence suggests that successful implementation will depend just as heavily on governance capability as on technical expertise.
Perhaps the most valuable takeaway is that effective policy rarely depends on finding perfect answers.
Instead, it depends on creating institutions capable of learning, adapting and improving as new evidence emerges. In an era where technological change often outpaces regulation, that ability may become one of government’s most important competitive advantages.
Source Information
Study Title: Digital Government Outlook 2026 – Adopting and Governing AI in Government
Authors: Organisation for Economic Co-operation and Development (OECD)
Journal/Publisher: OECD
Year: 2026
