Risk Management
Flowpay needed to speed expansion into new countries and improve lending efficiency because metadata generation for PSD2 transaction categorization and underwriting report preparation were manual, slow, and difficult to scale across markets.
Reported impact
Flowpay reduced the time needed to research and launch into new markets, cut recurring metadata cleanup work, and sped up underwriting so its small team can expand internationally with less manual effort.
Through IBM Client Engineering and the Fintechx program, Flowpay built two generative-AI-assisted workflows on IBM watsonx.ai to accelerate both market entry and underwriting. The first uses large language models plus internet search results to assemble an initial metadata dataset for each new country, helping Flowpay categorize PSD2 transaction data from prospective customers faster as it enters new markets. The second uses raw PSD2 bank-account transactions and calculated data points to generate a plain-language underwriting summary that describes an applicant’s financial health and lending risk, giving human underwriters a faster starting point. Both workflows were piloted with seven users over four weeks and run within an IBM Cloud-based data environment using supporting services such as object storage, Watson Studio, and PostgreSQL. Because the outputs are intended to be refined and exposed through APIs, Flowpay turned generative AI into a reusable part of its cross-market risk and operations stack rather than a one-off proof of concept.
Public web information and PSD2 transaction data flow into watsonx.ai-powered processes, one generates market metadata for new-country onboarding and the other produces underwriting summaries from bank transactions and calculated features, and the resulting outputs are refined for API delivery inside Flowpay’s lending platform.
“100% time savings for new market research and metadata creation”
“Both processes rely on IBM watsonx.ai”
“The first solution uses large language models (LLMs) to generate completely new content, and to process internet-based search results.”
“The second solution is designed to significantly speed up Flowpay‘s underwriting process.”
“~20% speed increase in loan approval process by under-writer”
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
IBM Sample Feb 13, 2026 BORICA AD BORICA needed to modernize national fraud prevention so it could stop fast, cross-channel attacks in milliseconds without disrupting legitimate payment traffic across multiple banks and service providers. 4,000+ transactions per second IBM Safer Payments IBM Sample Nov 26, 2025 Dollarito Dollarito wanted to lend responsibly to people with low or no credit history using richer behavioral and banking data, but it lacked the secure, compliant and scalable infrastructure needed for AI-powered credit… 40% Cost reduction due to automation in loan disbursement and collection process IBM Cloud for Financial Services · IBM watsonx IBM Sample Oct 30, 2025 Citi Citi wanted to modernize internal audit so thousands of auditors could review documents, assess risk and surface anomalies faster than was possible with manual methods. 2,500 auditors IBM Cloud Pak for Data · IBM OpenPages · IBM Watson Discovery 