Alice — Brazil's AI system for detecting procurement fraud
Brazil
Alice, deployed by Brazil's CGU since 2015, uses NLP and RPA to screen federal procurement notices daily, reducing audit turnaround …
Kazakhstan · Almaty · See the Kazakhstan profile
Kazakhstani anti-corruption NGO Adildik Joly uses Datanomix's Red Flags Management AI platform to flag risky public contracts, recovering KZT 244m in fines, 66 criminal cases and 1,100 administrative protocols since 2021, plus 170m tenge from an AI-detected VAT fraud scheme.
Kazakhstan built a centralised e-procurement portal in 2015, generating a dataset now exceeding 100 million rows of contract records. Since 2021, the anti-corruption NGO Adildik Joly — with 17 regional branches — has used this data to monitor public spending for fraud and waste.
Adildik Joly's tool is Red Flags Management (redflags.ai), built by the Kazakhstani software company Datanomix, which grants civil-society groups free access as part of its social-responsibility policy. The platform uses large language models to extract structured data from unstructured procurement PDFs, then screens transactions against 43 documented red-flag characteristics (e.g. a single bidder, prices far above market rate) to generate automated fraud alerts.
Documented results since 2021: KZT 244 million (~US$511,000) in fines, 66 criminal cases opened, and 1,100 administrative protocols issued against civil servants, alongside KZT 20 billion (~US$42 million) in high-risk purchases flagged. A 2022 case added a bespoke VAT-fraud-detection algorithm that traced illegitimate transfers to contractors, recovering 170 million tenge (~US$400,000) for the state budget. Kazakhstan's State Audit Institutions separately report identifying US$305 million in financial violations using related tooling, with about 70% of flags caught at the planning stage before money changes hands; Oxford Insights estimates potential annual savings of roughly US$86 million against the US$22 billion in procurement spend the system screens each year.
Caveat: these figures are self-reported by Adildik Joly and Datanomix rather than independently audited, and civil society's access to the underlying AI platform depends on a private vendor's voluntary grant — a arrangement that could be withdrawn, which is a real sustainability risk for a tool now central to Kazakhstan's anti-corruption monitoring.
Read the full analysis: https://www.open-contracting.org/2023/10/30/in-kazakhstan-opening-up-procurement-boosts-public-oversight-and-prevents-millions-in-wasteful-spending/
Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.
Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.
Where this practice's information was retrieved from, and when.
Brazil
Alice, deployed by Brazil's CGU since 2015, uses NLP and RPA to screen federal procurement notices daily, reducing audit turnaround …
Lithuania
Lithuania's STT, with the OECD and Government Transparency Institute (EU-funded), built a Hellinger Distance Random Forest on 11 merged datasets. …
Ukraine
DOZORRO's ML risk-flagging tool screens Ukraine's Prozorro procurement tenders for corruption red flags, catching 26% more unfair-selection and 298% more …
Romania
Romania's ANI automatically cross-references procurement officials' identities with registered bidders before contracts are signed; over 100,000 procedures worth RON 72 …
Open full copilot Grounded in cited practices — always check the sources.