Arva AI, a New York company that develops AI models and agents to automate high-stakes decisions for banks and fintechs, has closed a $20 million Series A round. Arthur Ventures led the investment, and existing backers Y Combinator and Gradient Ventures also participated.
The raise comes shortly after Arva launched its research division, Arva Research, in September, and follows a series of large deployments. The company counts a top-10 US bank, a payments institution in the S&P 500 and Wise among its customers.
Arva argues that banks still review most of their thousands of daily critical decisions by hand. It points to BCG estimates that AI agents could reduce bank costs by 30-40% by 2030, but notes that those savings depend on results that are reliable and explainable to regulators, a standard general-purpose models struggle to meet without human checks.
Arva Research focuses on building proprietary models for the decision steps where accuracy matters most. Its first model, Arva Intel, researches businesses and individuals online and, in the company’s own testing, beat leading general-purpose models on accuracy by 13%. It works alongside AgentCore, Arva’s continual learning infrastructure, which adapts to each bank’s procedures and policies. The new funding will go first to Arva Research and to extending the company’s agents across the back office, from financial crime compliance to lending.
“Banks still run most of their back office by hand. We’re building the models that let them hand that work to agents and trust the result,” said Rhim Shah, co-founder and CEO of Arva AI. Patrick Meenan, general partner at Arthur Ventures, said the firm was drawn to Arva’s domain expertise and its focus on accuracy and auditability.
Arva AI was founded in 2024 by Shah, who previously led financial crime product at Revolut Business, and CTO Oliver Wales, formerly Chief Product Engineer at Seal.