LocationsAI can help oversee public spending — but it also needs oversight
September 17, 2026
Artificial intelligence can make public audit more effective — but it must also become subject to audit itself. This was the key message delivered by Katarzyna Wawiernia, UNDP Resident Representative in Kazakhstan, at the international conference “Public Audit of the Future,” held to mark the 30th anniversary of the Supreme Audit Chamber of Kazakhstan.
She began by reflecting on how the very idea of accountability has evolved. The Sustainable Development Goals have pushed governments to look beyond whether public funds were spent lawfully, and to ask what real value those expenditures created for people, the economy and the environment. In other words, accountability has expanded from asking “where was the money spent?” to “what did that spending actually achieve?”
According to the Resident Representative, when countries try to answer this question, they often encounter the same pattern: the most serious risks in public administration tend to emerge at the intersection of institutions and sectors — where the responsibility of one entity ends and another begins. Health care offers a clear example. Reliable access to medicines depends on a full chain of decisions, from demand forecasting and financing to procurement, logistics and delivery to patients. When shortages occur, auditing only the final stage of delivery will rarely reveal the root cause, which is often linked to inaccurate demand planning or fragmented accountability across institutions. This is why audit must assess not only individual transactions, but the system as a whole.
The UNDP Resident Representative suggested applying the same logic to artificial intelligence. As a tool, AI is already helping audit institutions analyse large volumes of transactions, detect anomalies, identify risks in public procurement and spot duplicate payments, making audit faster and more effective. Yet as algorithms play a growing role in governance, budgeting and public service delivery, they too must become subject to audit.
She cited two international examples. In the Netherlands, the Court of Audit reviewed nine government algorithms in 2022 against requirements related to governance, data, privacy and ethics; only three met the basic criteria. In the United States, the Government Accountability Office developed a dedicated AI accountability framework covering governance, data, performance and monitoring. Auditors will increasingly need to ask what data an algorithm was trained on, how reliable it is, whether it can produce unbiased results, and who is accountable when an algorithm influences a public decision. “AI can automate analysis, but it must not automate accountability,” she said, emphasizing that responsibility for decisions must remain transparent.
Speaking about Kazakhstan, she noted that the country has already introduced legal requirements on data quality, lawful use and transparency of AI systems. The next step is to translate these principles into practical audit methodologies and institutional capacity. The UNDP Resident Representative proposed a three-tier model of public audit: financial control, performance control and algorithmic control. She also noted that UNDP stands ready to support this work, drawing on its expertise in public finance management, SDG financing and the responsible use of AI.
Related
content
content
