AI: A Game-Changer in Combating Corruption in Public Procurement
AI has the potential to revolutionise the fight against corruption in public procurement. It is providing governments with advanced tools to detect irregularities in real time and take swift and effective actions. Building on insights from an earlier blog by Oxford Insights, this article will take you through a number of AI-driven anti-corruption measures which are already making a difference globally.
In particular, we will highlight two groundbreaking applications: Ukraine’s Prozorro platform, which has saved billions of Euros by enhancing fraud detection; and Brazil's Alice tool, which has transformed procurement oversight with AI-driven automation.
Alongside this, we will be introducing our new LLM Company Analysis Tool, developed in collaboration with Oxford Insights. This AI-enabled tool aims to enhance procurement integrity by identifying trends within procurement data that are indicative of fraud. To ensure our tool is accessible and engaging, we produce an overall ‘trustworthiness score’ for procurements which provide the average reader with a quick, easy-to-understand overall assessment of a particular procurement.
Procurement Transparency in Ukraine
AI has played a transformative role in public procurement across Central and Eastern Europe (CEE), with a particularly notable example being Ukraine’s AI-powered Dozorro system. This system is integrated with Ukraine’s Prozorro platform, which processes over 3,000 tenders (formal trading contracts) daily. Dozorro has significantly enhanced fraud detection in Ukraine; since being paired with Prozorro, it has identified 26% more cases of unfair supplier selection and increased collusion detection by 298%.
How exactly does Dozorro do this? Dozorro is a feedback platform that is integrated with Prozorro and leverages AI tools and Machine Learning to monitor procurement activities thorough a public risk-based methodology. It identifies high-risk tenders and allows citizens, businesses, and civil society groups to submit complaints regarding procurement violations. This includes things such as non-competitive behaviour or poor-quality goods and services. This functionality allows users to flag potentially fraudulent or misleading contracts, enabling organisations to take action against procurement irregularities and fostering transparency and accountability in public spending.
The AI system then flags tenders with unresolved complaints or low satisfaction ratings for further investigation. By aggregating complaints into a single system. Dozorro facilitates quicker resolutions and promotes transparency, as all feedback is publicly accessible. This AI-driven platform shifts the anti-corruption architecture in public procurement from reactive to preventive. In turn enhancing accountability, preventing corruption, and ensuring better allocation of taxpayer money.
In collaboration with international partners in Europe, Dozorro has gathered a set of best practices for monitoring public procurement. These insights provide specialists from other countries with valuable guidance, allowing them to choose the most effective solutions by building on the successful approach Dozorro has taken to monitoring public procurement.
Brazil's Resistance to Adopting AI
Alice is an AI-driven tool developed by Brazil’s CGU to enhance public procurement oversight, focusing on combatting: fraud, errors, and inefficiencies. It automates the collection and analysis of procurement data from multiple platforms, using AI and other technical processes to assess tenders against 40 risk categories. It then generates tailored alerts, which enable rapid, preventive audits. This reduces audit time from over four hundred days to just eight.
As well as generating 1.3 billion Brazilian reais due to these kinds of operational savings, it has led to the suspension of potentially fraudulent tenders totalling 9.7 billion reais. The system's main advantages include enhanced transparency, more efficient public spending, and the prevention of corruption. However, its adoption is limited by technical challenges, such as data access and platform integration issues, and resistance to change within public institutions. This demonstrates a wider challenge with the application of AI-enabled solutions to procurement fraud.
Revolutionising Corruption Detection in Public Procurement: Our LLM Company Analysis Tool (LLM CAT)
Butterfly Data is proud to have developed our own AI-enabled tool to help fight against corruption. The LLM CAT, being developed in collaboration with Oxford Insights, assesses the integrity of companies that are involved in public procurement. It aggregates data from open sources such as Companies House, Contract Finders, Wikipedia, and other digital outlets and flags companies which demonstrate unusual attributes, or behaviours. It then assigns a ‘trustworthiness score’ to determine the overall safety of the information provided by the sources.
Public organisations looking to tender work can use this score to proactively detect and address potential ‘red flags’ in their public procurement journey. Trends such as disproportionate numbers of contracts being won by limited numbers of suppliers, or the persistent use of restricted procurement methods are identified through the use of LLM CAT; and the intelligence is made accessible for all. The UK does not currently have access to a standardised system that can aid in fraud detection, especially over multiple sources. This tool will be unique in its ability to provide a transparent and trustworthy approach to analysing procurement of public tenders. Moreover, this tool is accessible to a wide range of users, so can be adopted into several types of organisations.
An update on our progress with the tool will be revealed in the upcoming weeks.
If you have any questions on this article, or the tools mentioned, then feel free to ask me Maja Strawinska.