Violations and fraud in public procurement are usually identified only after the problem has already occurred. In Lithuania, however, a new approach aimed at detecting such risks in advance has been introduced. The Organization for Economic Cooperation and Development (OECD), in cooperation with Lithuania's Special Investigation Service, tested a system that uses artificial intelligence to identify risky cases in public procurement in advance. After the pilot phase was completed in 2026, the model moved to the stage of practical implementation and further refinement. The system analyzes data on public procurement, companies, their owners, and previously recorded violations, and assigns a risk level to each contract.
An important feature of the Lithuanian experience is that artificial intelligence does not replace people. It shows the inspector which transactions among thousands of contracts should receive attention first. A small number of tender participants, the repeated success of a single supplier, the systematic conclusion of contracts with a particular contracting authority, or the existence of ownership ties between companies may all be assessed as risk indicators. According to OECD trials, reviewing just a quarter of the contracts that artificial intelligence rated as the riskiest made it possible to cover nearly 60% of the violation cases.
This approach is also of practical importance for Uzbekistan. Public procurement in Uzbekistan is increasingly being digitalized, and large volumes of data are accumulating on electronic platforms and in state registers. Linking them through a single analytical system would make it possible to identify risks in the use of budget funds at an earlier stage. Yet there is no need to copy the Lithuanian model in full. As a first step, it would be advisable to implement a pilot project in selected sectors with high procurement volumes or among major state contracting authorities.
The pilot system could compare public procurement data with the register of legal entities, information on company founders, and other information databases that may lawfully be accessed, and generate a “risk index” for contracts. Tenders with low competition, the participation of interconnected companies, sharp price discrepancies, or repeat contracts with a single supplier would automatically be brought to an analyst's attention. This is not an automatic determination of guilt, but a tool for allocating inspection resources efficiently.
From an economic standpoint as well, such a system could be a convenient solution for Uzbekistan. The use of open-source software reduces substantial costs, while the main focus shifts to organizing state data and linking it together. With human oversight preserved, artificial intelligence helps to detect risky cases early, saves inspectors' time, and supports the efficient use of budget funds.
Source: OECD, “AI-driven fraud detection and digital transformation in law enforcement in Lithuania”
Alisher Azizov Sustainable Development Center — Chief Specialist