08/06/2026
Artificial intelligence is beginning to transform public auditing: from anomaly detection to predictive auditing
by Iván Martínez, CEO of Intedya
Artificial intelligence is beginning to bring about significant changes to the way in which public audit institutions carry out their control, oversight and accountability functions. Although its adoption is still in its early stages, the progress is clear: an increasing number of audit bodies are incorporating AI tools to analyse large volumes of data, detect anomalies, prioritise risks, process documents and improve the efficiency of their audit work.
This is reflected in the OECD report *The State of Artificial Intelligence in Public Audit. Evidence from selected countries and the European Union, published in May 2026 as part of the OECD Artificial Intelligence Papers series, which analyses the state of artificial intelligence in public audit based on consultations carried out between March and July 2025 with 15 institutions from 14 countries and the European Union.
The document reveals a situation that remains uneven, but with a clear trend: public audit is gradually shifting away from traditional models, based on manual reviews and limited sampling, towards more digital, preventive and data-driven approaches.
According to data collected by the OECD, two-thirds of the institutions surveyed already have a formal artificial intelligence strategy in place. Furthermore, 80 per cent have internal guidelines or policies on AI, 87 per cent provide training for staff, and a further 87 per cent report having at least one AI tool in production. These figures show that artificial intelligence is no longer an experimental or purely technological issue, but has become a strategic component of the digital transformation of public audit.
One of the areas with the greatest potential is the detection of anomalies and risk assessment. Audit institutions are using machine learning models to identify unusual transactions in accounting records, procurement files, payments, grants or administrative databases. These tools enable the analysis of entire data sets, rather than relying solely on small samples, and help to direct audit resources towards transactions, entities or processes most likely to involve irregularities.
Public procurement stands out as one of the most significant areas. Given its economic scale, complexity and exposure to risks of fraud, collusion, bribery or conflicts of interest, the application of AI can substantially improve oversight capabilities. The report highlights examples such as that of Costa Rica, where anomaly detection algorithms have been applied to data from the national electronic procurement system to identify unusual patterns and strengthen the control function. Predictive and preventive models are also highlighted. Rather than an audit focused solely on detecting problems after they have occurred, some institutions are beginning to use AI to anticipate risks. The case of Italy is particularly illustrative: its Court of Auditors has developed an experimental model to predict the risk of financial difficulties in local authorities, combining financial, demographic and institutional data. Tools of this kind make it possible to direct supervisory action at an earlier stage and take corrective measures before problems become more serious.
Another significant application is intelligent document processing. Public audit deals with large volumes of contracts, invoices, supporting documents, reports, decisions and administrative files. AI enables the extraction of structured data from unstructured or semi-structured documents, reducing repetitive manual tasks and facilitating more comprehensive analysis. Technologies such as OCR, natural language processing and language models enable documents to be classified, amounts, dates, names, concepts or regulatory references to be extracted, and scattered documentation to be converted into analysable information.
Generative AI is also beginning to gain ground, albeit with greater caution. Its use is primarily focused on support tasks: summarising lengthy documents, preparing drafts, translating texts, improving communications or facilitating internal searches. Under no circumstances is it intended as a substitute for the auditor’s professional judgement. The report emphasises that the institutions using these types of tools maintain processes involving human review, quality controls and validation by experts.
One of the most interesting examples is that of the Cour des comptes in France, which has developed its own generative assistant, designed for tasks such as summarising, translating, rephrasing, document analysis and research support. The key to this approach lies in technological sovereignty and security: the tool has been designed to maintain control over data, protect confidentiality and comply with applicable legal requirements.
However, the OECD report also warns of a significant gap between pilot projects and large-scale operational roll-out. Many institutions are experimenting with AI, but not all have the infrastructure, data, in-house expertise and governance frameworks needed to integrate it seamlessly into their audit processes.
Data governance emerges as the key critical factor. Without comprehensive, interoperable, standardised, accessible and secure data, artificial intelligence loses its reliability. Fragmented systems, poor data quality and a lack of common standards hinder the scalability of solutions and may undermine confidence in the results. In auditing, where conclusions must be traceable, explainable and defensible, this issue is particularly sensitive.
There are also significant challenges relating to ethics, explainability, bias, security and reliance on third parties. Public audit institutions cannot simply adopt AI tools for the sake of efficiency. They must ensure that their use respects the principles of transparency, accountability, confidentiality, independence and public trust. AI can help prioritise risks or detect patterns, but the final decision must remain with competent auditors.
The report also highlights an important conclusion: the institutions making the most progress are not necessarily those that purchase the most technology, but those that combine leadership, governance, training, staff involvement and in-house development capacity. AI in public audit is not merely an IT project. It is a methodological, organisational and cultural transformation.
From the perspective of compliance and public oversight, this evolution has far-reaching implications. Public audit can enhance its preventive capacity, improve the detection of irregularities, broaden the scope of its reviews and generate early warnings regarding financial, contractual or integrity risks. But to achieve this, it needs to build a solid foundation: reliable data, technical expertise, ethical standards, quality controls and appropriate oversight frameworks.
Artificial intelligence will not replace the public auditor, but it will change the nature of the audits that institutions can carry out. The future points towards audits that are more continuous, more predictive, more data-driven and more risk-oriented. The challenge will be to ensure that this technological evolution reinforces, rather than undermines, the core values of the audit function: independence, objectivity, evidence, transparency and public trust.
Ultimately, AI presents a historic opportunity to modernise public audit. Its true value will lie not in automating isolated tasks, but in enhancing institutions’ ability to anticipate risks, detect irregularities and better protect public resources.
Data source: OECD, *The State of Artificial Intelligence in Public Audit. Evidence from selected countries and the European Union*, OECD Artificial Intelligence Papers, No. 58, May 2026.
