DataVisor, which provides a real-time AI decisioning platform for fraud and risk prevention, unveiled the capability on the Money20/20 USA Breaking News stage. Vera Code is intended to reduce the time and complexity involved in building, adapting, and managing financial crime programmes. According to the company, the product also shifts AI from handling isolated tasks to agents that run programmes end-to-end, with humans remaining in control.
Addressing implementation complexity
DataVisor describes fraud as having become polymorphic. Attackers now deploy AI agents that adapt, coordinate, and scale at machine speed, which causes static legacy defences to lose effectiveness more quickly. The company also notes that institutions often adopt newer solutions slowly because implementation appears risky and complex. Once those solutions are live, programmes struggle to keep pace with changing threats, risk signals, and regulatory requirements. In an exclusive interview with The Paypers, held at Money20/20 USA, Patrick Harr, CEO of DataVisor, said: `Agentic fraud is nothing more than coordinated fraud on steroids. It’s moving at machine speed, the sheer number of attacks is growing, and the agents can coordinate and adapt on the fly.`
Much of the burden lies in configuration. Fraud and AML teams must turn risk policies and operational requirements into data, detection, and workflow configurations, and then repeat that work with every update. With this in mind, DataVisor pairs its out-of-the-box fraud and AML solutions, built on more than ten years of experience, with Vera AI agents that tailor each programme and refine it as threats change. Patrick Harr also told The Paypers that `Even on the anti-money laundering side, real-time is becoming more important. You can automate SAR filings and investigation summaries, and use those signals to feed back into fraud.`
How Vera Code works
Users work with Vera Code through a conversational interface. The tool takes customer requirements and builds either a complete solution or individual components. These include event schemas, aggregated signals, rules, scorecards, alert queues, review workflows, and test data. Each build is tested before going live. Later changes, made as requirements or threats evolve, are verified and approved before they take effect.
Maxim Spivakovsky, Sr. Director, Global Payments Risk Management, SoFi Tech Solutions, said the agents allow fraud teams to apply their own data and expertise to develop strategies faster and manage programmes more efficiently. In addition, Jim Mortensen, strategic advisor in the Fraud and AML practice at Datos Insights, said they support a new operating model in which practitioners build detection, decisioning, investigations, and reporting through natural language.
Speaking with The Paypers on the launch, Patrick Harr said: `I was going to call it a full fraud defence, but I’d rather characterise it as a fraud offence. We can now implement it in hours, as opposed to days, weeks or months.`, adding that `Low-code, no-code still required data science knowledge. This is truly as simple as a ChatGPT interface. You can type ‘lower my false positives but keep my same precision’, and it gives you a full strategy. It’s still human in the loop, with full transparency for the auditors. Banks and payment processors want to get a lot more efficient. They don’t necessarily want to hire more people. If you can make your current people more productive, that’s a significant benefit.`
Early results and the wider platform
DataVisor reports that Vera Code builds new end-to-end solutions up to 100 times faster and updates existing ones up to 25 times faster. The company delivered a complete real-time fraud solution for a Fortune 500 insurer in four days, against an industry norm of three to six months. Furthermore, a regional bank replacing its legacy system had budgeted for an implementation lasting several months. With Vera Code, it is saving hundreds of thousands of USD through simpler implementation and management.
Vera Code is part of the broader Vera offering, which combines conversational AI with execution agents across detection, optimisation, investigation, and reporting. The agents run on DataVisor’s governed platform. The platform aggregates and enriches signals and data and makes real-time decisions faster, while also applying adaptive AI, including patented unsupervised machine learning designed to uncover emerging and coordinated fraud. Agents can adapt programmes using first- and third-party data alongside investigation findings, while human oversight is retained. Adding on this, the company told The Paypers that `Hedge funds on Wall Street make their money on milliseconds. I believe fraudsters will do the same thing. That’s why fraud protection has to shift from defensive mode to offensive mode.`
For institutions in regulated environments, the requirement that every change be verified and approved is central to the proposition. It allows configuration work that has traditionally involved implementation cycles lasting months to be automated without removing the oversight that regulators expect.
