TPBank Wins AI Hackathon 2026 With an Agent That Cuts Credit File Checks From 45 to 10 Minutes
Tien Phong Commercial Bank (TPBank) has won AI Hackathon 2026, a contest held during the Smart Banking 2026 conference, with an AI assistant that processes corporate credit applications, Vietnam Finance reported in coverage republished by Stockbiz on October 23. The bank's TPB Trailblazers team beat 14 teams from financial institutions and took first place in the final on October 22.
The contest was organised by the State Bank of Vietnam's IT department, the Vietnam Banks Association and IEC Group, and asked teams to build AI applications on modern agent platforms. TPBank's entry combines agentic document extraction with a multi-agent design: several independent AI agents coordinate, extract, cross-check and standardise data from the large bundles of documents that business loan appraisals require. The bank says the tool cut the time to process one credit file from 45 minutes to 10, with accuracy of up to 91 percent.
The team was assembled from staff across departments, had only days to prepare and began building the product a week before the competition. The judges described the solution as highly practical and said it could be applied immediately to the bank's existing data systems.
What it means in practice
Credit appraisal is one of the most paper-heavy jobs in Vietnamese banking, and cutting manual reading and re-keying could shorten loan turnaround for small and medium-sized businesses. But a hackathon result is not a production system, and accuracy of up to 91 percent still leaves room for errors, so the assistant would need human review before any lending decision, in line with the State Bank's risk tiers and human review rules for banking AI. The win also fits a wider push by lenders and regulators, including the central bank's search for a developer of an AI credit platform and Viettel's AI platforms for banks.
TPBank has not said when, or whether, the assistant will move into live credit operations, how its accuracy was measured, or how customer data would be protected when agents process loan documents.
What to watch: a production rollout timeline at TPBank, whether other banks bring similar document agents into appraisal, and how the State Bank's AI rules are applied to agent-based credit tools.