AI in Financial Services

Revolut, the London-based digital bank, self-hosts an open-weight GLM coding model on 56 NVIDIA B200 GPUs to give about 2,000 engineers flat-cost, unmetered AI coding. Its AI operations lead reports a 60% cut in coding time and 70% in lead time after the switch in April 2026, a blended cost of about $0.30 per million tokens, roughly one tenth of its previous frontier-lab spend, and an 88% latency improvement from inference tuning on the same hardware. A sandbox with operation allow-lists keeps a human in the loop for EU banking supervision.

Czech fintech Patron GO (Antivirus s.r.o.), with 600,000+ users, is building Patron GO 2.0 with BigHub and Czech Technical University: ML models that categorise bank transactions automatically, community self-labelling that retrains them, peer benchmarking, and generative-AI personalised advice. The EU-co-funded project (June 2025 to May 2027) is in a Czech pilot, with Slovakia and Poland as the first foreign markets.

Satispay, Italy's independent payment network, rolled out Claude Code to every engineering laptop after a 30-day evaluation against another AI coding tool. More than 75% of code committed each month is now generated with Claude, adoption is above 90% of engineers, and the Java 8 to Java 21 upgrade of the core transaction service took under four days against a four-week estimate.

Trustly, a Swedish open-banking payments company, moved its analytics to Amazon Redshift and Amazon QuickSight and added Amazon Q in QuickSight for natural-language questions. Users grew from 80 to 300, dashboards refresh every 15 minutes instead of three times a day, and over six months staff asked 15,000 questions with 89% accuracy; AWS reports insights 4x faster.

Revolut is rebuilding its AI stack around PRAGMA, a family of transaction foundation models trained on NVIDIA H100 GPUs through Nebius AI Cloud, to replace separate fraud, credit, engagement and recommendation models with one shared behavioural layer. NVIDIA reports 3-5x faster model development and 2-5x higher training throughput, and Revolut's internal benchmarks show 65% more fraud caught.

African fintech Chipper Cash replaced its third-party KYC service with an in-house facial similarity system backed by Pinecone vector search, cutting new-user verification latency from up to 20 minutes to under 2 seconds and reducing fraudulent sign-ups by 10x across its 7-country African market, with Uganda as the representative case.

Miden deployed an AI-powered transaction monitoring tool on AWS infrastructure that reduced fraud detection time by 82%, improved processing scalability by 75%, and cut manual monitoring effort by 67% across its Africa payments and banking operations.

Visa's AI-driven fraud detection framework combines real-time threat monitoring, continuous ML model training, customizable rules engines, and AI-enhanced malware detection across its global payments network — balancing fraud prevention with seamless customer transactions.