OECD finds audit institutions are building AI capacity but struggling to scale

Public audit institutions are expanding their use of AI, but most remain at an early stage of adoption, with a significant gap between pilot projects and full operational deployment, according to a new OECD paper.

Drawing on consultations with 15 institutions across 14 countries and the European Union, the paper says AI is being explored to strengthen oversight and improve audit processes in areas such as anomaly detection, document processing, knowledge management and predictive risk assessment.

The OECD says institutional commitment is already visible across several indicators. Among the institutions consulted, 67% reported having a formal AI strategy, 80% had internal AI guidelines or policies, 87% offered AI-related staff training, and 87% had at least one AI tool in production.

However, the paper stresses that maturity levels vary widely and that many tools remain limited in scale or are still being tested. It identifies a gap between experimentation and scalable operational deployment, despite the growing integration of AI into broader digital transformation efforts.

The paper highlights several emerging audit use cases, including machine-learning systems for anomaly detection in procurement and financial records, predictive models to identify entities at higher risk of distress or non-compliance, intelligent document processing for extracting data from unstructured files, and generative AI tools for drafting, summarising and translating documents.

It also points to more specialised applications, such as semantic search, knowledge management, and visual or spatial analysis using satellite imagery, drones or other sensor-based systems.

Despite growing experimentation, the OECD says the main barriers to wider use remain structural. Fragmented data systems, weak interoperability, limited internal technical expertise and uneven digital infrastructure continue to slow progress.

The paper argues that robust data governance, secure and interoperable systems, and stronger in-house development capacity will be critical if public audit bodies are to scale AI responsibly while maintaining transparency, accountability and public trust.

It also stresses that AI is being positioned as a support tool rather than a substitute for auditors. Across the cases reviewed, human oversight remains central, both because of current limitations in explainability and reliability and because audit institutions are treating AI adoption cautiously in high-stakes oversight settings.

The OECD presents the current period as a transitional phase in which public audit institutions are building the foundations needed for broader and more trustworthy use of AI in oversight.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!