Opening a bank account means passing a know-your-customer (KYC) check. The bank also has to make sure the customer is not on a sanctions list and that their transactions look legitimate. Together with anti-money-laundering (AML) controls, this is financial crime compliance, and at most banks it generates thousands of decisions a day, many of which are still reviewed by people one by one. Arva AI, a U.S. startup that wants AI agents to take that work over, has raised a $20 million Series A.

Arva AI builds AI models and agents that automate repetitive KYC and AML work at banks and fintechs, along with lending decisions. Its main products cover onboarding checks for individuals and businesses (KYC and KYB), sanctions screening, suspicious-transaction investigations, and deeper due diligence on higher-risk customers. According to the company, it has deployed at scale over the past year with a top-10 U.S. bank, an S&P 500 payments institution and Wise. The first two are unnamed. Its website says it processes more than a million reviews a month.
The key issue is accuracy. A bank’s decisions have to be explainable to regulators, and the company argues that general-purpose AI models are not accurate enough and often cannot show how they reached a conclusion. It cites BCG’s estimate that AI agents could cut bank costs 30% to 40% by 2030, but says that holds only if results can be trusted and explained.
That is why the first use of the new money is Arva Research, the research unit it launched in September. It builds 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%. The company has not published the test data or method. It is paired with AgentCore, infrastructure that keeps learning from a bank’s feedback so the models adapt to each bank’s operations and policies.
Co-founder and CEO Rhim Shah previously led financial crime product at Revolut Business. Oliver Wales is co-founder and CTO, and the two founded the company in 2024.
Arva announced the round. Arthur Ventures led it, with continued participation from Y Combinator and Gradient Ventures, Google’s early-stage AI fund. According to press reports, the company raised a $3 million seed round led by Gradient in January 2025. It did not disclose a valuation.
The money goes first to expanding Arva Research, then to extending agents across the bank back office, from financial crime compliance to lending. The company also named disputes and payment operations. Patrick Meenan, general partner at Arthur Ventures, said the firm backed Arva for its domain expertise, technical capability and commitment to accuracy and auditability.
The market is already crowded. Rule-based monitoring systems have long been plagued by false positives, which Wowtale’s regtech landscape puts at more than 90% of alerts, and AI agents are moving in. Variance, a Y Combinator startup founded in 2023 by former Apple fraud engineers, runs agents for KYC, KYB and AML investigations and transaction monitoring, and raised a $21.5 million Series A in March led by Ten Eleven Ventures. London- and Berlin-based Diligent AI automates KYC and AML alert handling and raised a $2.5 million seed round led by Speedinvest and Shapers. Denmark’s spektr raised a $20 million Series A led by NEA in April, according to press reports, to automate KYC for businesses.
There are established players too. Munich-based Hawk, founded in 2018, combines payment screening, AML transaction monitoring and ongoing KYC on one platform for more than 80 financial institutions, and raised a $56 million Series C led by One Peak last April. New York’s Sigma360 integrates risk data and screening technology and raised a $17.3 million Series B led by Moderne Ventures in March, while London’s ComplyAdvantage supplies sanctions, PEP and adverse-media data in real time to more than 1,000 customers. For more on the competitive landscape, see here.
Arva’s pitch is that building its own models for each step of the work will make agents accurate enough for banks to trust. The 13% improvement is the company’s own figure, though, and how much of a bank’s work can safely be handed to agents will depend on the evaluation results and customer deployments it shares next.
MORE FROM THE POST
- Model ML Raises $250M at $1.5B Valuation for Personal AI Agents in Finance
- Spend Management Platform Payhawk Raises $280M Series C at $2B Valuation
- Infisical Raises $40M Series B at $600M Valuation to Secure Credentials for AI Agents
- Dealership AI Agent Flai Raises $27M Series A
- Raindrop Raises Series A, Bringing Total Funding to $50M
- agentic AI
- AI
- Arthur Ventures
- Arva AI
- compliance
- Fintech
- funding
- Gradient Ventures
- Overseas funding
- regtech
- seriesA
- Y Combinator
Share
Most Read
- 1
- 2
- 3
- 4
- 5


Leave a Reply