Latvian Scientists Create A Tool For More Efficient Public Procurement Evaluation

Much of public procurement evaluation consists of manual work by experts and specialists. This is time-consuming, since it involves comparing extensive documentation, evaluating compliance with regulatory requirements—like whether there are competition-restricting criteria—and other factors. The assessment criteria alone can stretch to several dozen pages. It is a systematic process that often takes two specialists several days. Taking these factors into account, as part of a national research programme project, the Central Finance and Contracting Agency tasked researchers with assessing whether AI could reduce the manual workload and increase the efficiency of the evaluation process.

The national research programme project “Analysis of the Suitability of Artificial Intelligence Methods in the Field of European Union Fund Projects” seeks solutions as to how generative AI can help to analyse European Union-funded project procurement documents, in order to significantly improve the efficiency and quality of work of procurement specialists in both the public and private sectors.

Under the aegis of this project, scientists from the University of Latvia have created and tested a new prototype of generative AI technology that can significantly ease part of the procurement review workload and improve the quality of procurement oversight in public administration. This solution tackles existing problems in procurement evaluation: an initial assessment can be carried out quickly and objectively for all procurements, while specialists can devote their expertise to more complex assignments and to regulatory aspects where generative AI is not yet able to provide sufficiently high-quality and fully reliable assessments. This can not only improve efficiency and quality and reduce errors, but also reduce manual document processing and the need for repeated review steps.

Routine work entrusted to AI
The AI tool prototype developed by the scientists is intended to automate the first round of evaluation of European Union fund project procurement documents, whereby the system reads and segments procurement regulations and draft contracts, links them to the Public Procurement Law, and analyses them according to several dozen predefined criteria. Experts then focus on more complex assignments instead of spending time on more routine tasks that can be entrusted to AI.

“It is an AI tool that can help the Central Finance and Contracting Agency monitor procurements. Accordingly, AI analyses documents and seeks answers to questions such as whether competition is being unjustifiably restricted. It also checks whether procurement rules comply with laws and other regulations and are transparent,” explains the project’s co-author, scientist Raivis Skadiņš.

In developing this AI tool for the public sector, the scientists’ main task is to use it correctly—assigning routine, systematisable work to the technology, while leaving value-added tasks to humans, where their powers of interpretation and contextual understanding are genuinely required. The benefits of this transition from manual reviewing to risk management-based in-depth additional checks include overall improvement in the quality of public administration work and significant time savings.

Decision-making remains in human hands
The AI tool prototype has been developed and adapted to assist the Central Finance and Contracting Agency in its work supervising projects financed by foreign financial instruments.

“Although AI cannot fully replace the work of our specialists, it can be useful. We have found that about 70% of the review questions for a specific procurement procedure on which the generative AI system was tested can be handled by the system at the initial evaluation stage. This means that tasks in which expertise, understanding, and decision-making are required are still performed by humans, while routine tasks are handled by AI,” says Anita Krūmiņa, Director of the Central Finance and Contracting Agency.

Thanks to its open-source code, this solution can also be used and further developed by other state and municipal institutions where large volumes of similar documents or projects need to be evaluated and supervised. “This prototype is open and accessible to everyone—other institutions and companies included. It is a shared resource on which future digital governance initiatives can be built,” adds Krūmiņa.

Also suitable for use by other institutions
The Central Finance and Contracting Agency is not the only institution in Latvia with a legal compliance review mandate. The Procurement Monitoring Bureau, the Ministry of Environmental Protection and Regional Development—for example, in the evaluation of municipal binding regulations—the administrators of funding projects at the Latvian Council of Science, the State Audit Office, and other entities that regularly assess frequently changing documentation have similar requirements. All these bodies are confronted by large volumes of documents, dispersed regulatory frameworks, and the imperative to ensure consistency and continuity.

Therefore, the scientists have published datasets and open-source tools that allow other institutions or companies to adopt them. Therefore the private sector can offer tailored solutions and services based on the platform built by the academic sector. Such solutions not only result in more efficient governance, but also in new forms of cooperation between the state and business via shared use of data and technology.