Vera Code is DataVisor's AI agent for building and updating complete fraud and AML programs. Teams describe what they need in plain language—a new ACH fraud solution, a ruleset aligned to new regulatory guidance, an investigation checklist based on an SOP—and Vera Code builds the working system: data transformations, rules, scorecards, decision flows, alert queues, case workflows, dashboards, and reports. Every change is presented for approval before it goes live, with auditability and governance built in. The result is a complete fraud and AML program deployed in days instead of months, and refined in minutes after that.
Vera Code launches today. It extends Vera, DataVisor's AI agent for fraud and AML teams, from improving individual detection components to building and evolving entire programs. This article covers what Vera Code is, how it works, what it builds, early customer results, and what to look for when evaluating AI agents for fraud and AML.
For fraud and AML leaders, the significance is speed with control. Financial crime changes faster every quarter as attackers use AI to adapt and coordinate. The institutions that pull ahead will be the ones whose defenses change just as fast, and whose experts spend their time on strategy instead of configuration.
What Is Vera Code?
Vera Code is the solution-building capability of Vera. It's designed for the people who build and run fraud and AML systems—platform administrators, system integration engineers, and fraud and AML solution builders—and it works at the level of the whole program rather than a single rule.
A traditional deployment runs separate work streams for data mapping, rule authoring, scorecard design, decision logic, case configuration, and reporting. Vera Code handles all of them from a single conversation. Teams bring their intent and their inputs—a requirements document, an SOP, a regulatory bulletin, or one of DataVisor's out-of-the-box solutions—and Vera Code turns them into a working, tested configuration on the DataVisor platform.
Vera now works in two modes, each built for a different job:
The short version: Vera Chat improves components. Vera Code builds and updates the system.
Why Fraud and AML Programs Have Taken Months to Build
Launching a new fraud or AML use case has traditionally been a multi-month project. The industry norm for deploying a real-time fraud solution is three to six months, driven by a sequence of manual steps:
- Data mapping and transformation to bring source data into the platform's schema and validate its quality
- Manual rule authoring across features, rules, and scorecards, often translated by hand from a requirements document
- Decision logic and case configuration built screen by screen, then wired together
- Vendor and engineering cycles for anything outside a self-service interface
- Testing and iteration across each module before go-live
Each step is reasonable on its own. Together, they create a long path between knowing what a program should do and having a program that does it. And the work doesn't end at launch. New typologies, new regulatory guidance, and new business lines each trigger another round of tickets and configuration.
The pace of financial crime makes that path matter more every year. AI has made attacks polymorphic: they adapt, coordinate, and scale at machine speed. Meeting that pace calls for defenses that adapt just as quickly. DataVisor's approach centers on three capabilities working together—perceive the full picture by connecting every relevant signal, adapt as new patterns emerge, and decide in milliseconds at scale. Vera Code is the newest part of the adapt layer, and it starts where adaptation has historically moved slowest: building and changing the program itself.
How Vera Code Works: From Intent to Working System
Vera Code follows the same pattern for every build, whether it's a new end-to-end solution or a targeted update to an existing ruleset.
Step 1: Describe the outcome
Teams state what they want in plain language and attach the relevant inputs: a requirements document, an investigation SOP, a regulatory bulletin, or an existing DataVisor solution to start from.
Step 2: Review the plan
Vera Code proposes a design before it changes anything—the rule conditions it would add, for example, or the components a new solution will include. It flags missing data and undefined parameters, like a time window a bulletin doesn't specify, so the team can resolve them up front.
Step 3: Build across every module
Once the design is approved, Vera Code builds the connected components: data transformations that conform to the DataVisor schema, features, rules, scorecards, decision flows, alert queues, case workflows, checklists, and dashboards.
Step 4: Test end to end
Teams can follow a test transaction from decisioning to an alert and into the investigation view, confirming the components work together before go-live.
Step 5: Deploy and keep refining
After launch, the same conversation becomes the way the program evolves. Teams refine strategies, update investigation workflows, and make controlled bulk updates in minutes as threats and business needs change.
What Vera Code Builds
Vera Code covers the full fraud and AML program lifecycle. Each capability below works on its own or as part of a complete build.
End-to-end fraud and AML solutions
Launch a new use case with every component configured and connected: data, detection, decisioning, and investigations. Starting from a DataVisor out-of-the-box solution and a requirements document, Vera Code produces a program shaped to the institution's specific needs.
Data integration and transformation
Map source data into the DataVisor schema, run data quality tests, and connect third-party data, so the signals your team needs are available to rules, models, and investigators.
Rulesets, scorecards, and decision flows
Create and update detection logic and the decision flows that act on it. Vera Code can compare an existing ruleset against new guidance, propose changes, and implement them once approved.
Alert queues and case workflows
Design queues, routing, and case workflows that match how your team investigates, with queue-specific configuration for different alert types.
Investigation checklists from SOPs
Turn written investigation procedures into embedded, interactive checklists that live inside the alert queue. Checklists are dynamic: they follow the SOP's decision tree and open follow-up questions only when the answers call for them.
Dashboards and reporting
Build BI dashboards and program reporting from a prompt, then adjust them with follow-up requests, such as switching a monthly trend to weekly or adding a new metric.
Controlled bulk updates after go-live
Apply changes across many components at once, with review and approval, so large updates stay consistent and auditable.
Vera Code in Action: Four Use Cases
These examples reflect the requests fraud and AML teams bring to Vera Code. Each starts with a plain-language prompt.
1. Launch a new ACH fraud solution
The goal: Onboard a new ACH fraud use case that reflects the institution's specific requirements.
“Build an ACH fraud solution taking into account the requirements document. The solution should include all the end-to-end components, including features, rules, scorecards, decision flows, investigation queues, and a monitoring dashboard.”
What Vera Code does: Starts from DataVisor's out-of-the-box ACH solution, transforms the data to the DataVisor schema, applies the customer's requirements, and builds every module. The team then traces a test transaction through the decision flow into an alert and investigation page to confirm everything works together.
2. Update an AML ruleset with new regulatory guidance
The goal: Close coverage gaps after FINTRAC published indicators of extortion-related money laundering.
“Compare this FINTRAC bulletin with my existing AML ruleset. Propose changes covering customer profile, multi-branch cash deposits, incoming transfers, and international wires. Flag missing data or undefined time windows, and show the design for approval before changing my ruleset.”
What Vera Code does: Maps the bulletin's indicators to the current ruleset, proposes condition groups for each gap, and flags parameters that need a decision. Numeric thresholds remain the institution's implementation choice. After approval, Vera Code implements the update, and the team can open a triggered alert to see exactly which conditions fired.
3. Turn an SOP into an investigation checklist
The goal: Make wire fraud investigations faster and more consistent.
“Create an investigation checklist for our Wire Fraud alert queue using this SOP. Include the required questions, and show additional questions when necessary.”
What Vera Code does: Turns the SOP into an embedded, dynamic checklist inside the Wire Fraud queue. The checklist follows the SOP's decision tree, opening follow-up questions only when the answers call for them, so every analyst works the same procedure, every time.
4. Build a program performance dashboard
The goal: Get a recurring view of program performance without assembling reports by hand.
“Build a BI dashboard for my ACH fraud program. Include transaction count, transaction fraud rate, transaction alert rate, alert volume monthly trend, false positive rate, and a trend graph showing fraud loss prevented by approach, both alert review and auto decline.”
What Vera Code does: Produces the first version of the dashboard, then updates it from follow-up prompts, such as changing alert volume to a weekly trend or adding a graph of the auto-decline rate.
Early Results: From Months to Days
Early results show what changes when AI handles the build. Vera Code builds new end-to-end solutions up to 100 times faster and updates existing ones up to 25 times faster.
Customers are seeing the same shift in production:
A Fortune 500 insurer went live with a complete real-time fraud solution in four days. The industry norm for this kind of deployment is three to six months of manual rule authoring, vendor cycles, and engineering work. With Vera Code, the full solution—rules, decisioning, and investigation workflows—was delivered in four days.
A regional bank is saving hundreds of thousands of dollars. Replacing its legacy system, the bank had budgeted for a months-long implementation while evaluating vendors. With Vera Code, it is saving hundreds of thousands of dollars through more streamlined implementation and ongoing management.
“Vera AI agents help fraud teams use their data and expertise to develop strategies more quickly and manage fraud programs more efficiently. DataVisor makes AI capabilities more accessible and practical, helping financial institutions like SoFi Tech Solutions adapt to evolving requirements and emerging fraud threats.”
— Maxim Spivakovsky, Sr. Director, Global Payments Risk Management, SoFi Tech Solutions
“Vera AI Agents represent a leap forward in financial-crime technology. They allow practitioners to create a fundamentally new operating model by turning intelligence into governed action across the entire fraud and AML lifecycle. Teams can build detection, decisioning, investigations, and reporting through natural language, enabling faster response to evolving threats.”
— Jim Mortensen, Strategic Advisor, Fraud and AML Practice, Datos Insights
Here's how the work changes across the program lifecycle:
Built for Control: Governance, Auditability, and Human Approval
Speed only matters if teams can trust what gets deployed. Vera Code is designed so the institution stays in control at every step.
- Design before change. Vera Code shows its proposed design for approval before modifying a ruleset or workflow.
- Gaps flagged, not guessed. Missing data, undefined time windows, and open parameters are surfaced for the team to decide.
- Testable outcomes. Teams follow test transactions end to end and review triggered alerts to see which conditions fired.
- Auditability and governance built in. Changes are reviewable and traceable, supporting model governance and examination readiness.
- Your expertise, Vera's execution. Teams set strategy and thresholds; Vera Code does the configuration work.
This matters for fraud and AML programs in particular. Regulators increasingly evaluate programs on demonstrated effectiveness, and every detection change needs a clear rationale. A model where people define intent and approve the design, while AI handles execution, keeps that rationale explicit and on the record.
Where Vera Code Fits in the DataVisor Platform
Vera Code builds on DataVisor's real-time fraud and AML platform, which is designed to stop financial crime in the blink of an eye through three capabilities:
Programs Vera Code builds run on the same real-time decisioning engine, with one data layer feeding detection, investigation, and reporting. That's what makes end-to-end builds possible: there's one platform to configure, not a collection of tools to integrate.
Vera Code also joins a set of Vera capabilities that already span detection, optimization, investigation, and reporting. Across that lifecycle, Vera has saved customers an estimated 1.7 years of work, cut investigation time by up to 60%, reduced regulatory reporting time by up to 90%, and lowered false positives by 87%.
What to Look for in AI Agents for Fraud and AML
AI agents are fast becoming a standard part of fraud and AML technology evaluations. These criteria help separate agents that change how programs are built from assistants that only answer questions.
Frequently Asked Questions
What is Vera Code?
Vera Code is DataVisor's AI agent capability for building and updating complete fraud and AML programs. Teams describe what they need in natural language, and Vera Code builds the working system—data transformations, rules, scorecards, decision flows, alert queues, case workflows, dashboards, and reporting—on the DataVisor platform, with approval steps, auditability, and governance built in.
How is Vera Code different from Vera Chat?
Vera Chat helps business and fraud analysts build and improve individual detection components, such as features, rules, and lists, and analyze their performance. Vera Code works at the program level, building and updating complete solutions end to end. Vera Chat improves components; Vera Code builds and updates the system.
How long does it take to deploy a fraud solution with Vera Code?
Vera Code builds new end-to-end solutions up to 100 times faster and updates existing ones up to 25 times faster. One early customer, a Fortune 500 insurer, went live with a complete real-time fraud solution in four days, against an industry norm of three to six months.
Does Vera Code make changes without human review?
No. Vera Code presents its proposed design for approval before modifying rulesets or workflows, and flags missing data or undefined parameters for the team to resolve. Teams stay in control of strategy and thresholds.
Who should use Vera Code?
Vera Code is built for platform administrators, system integration engineers, and fraud and AML solution builders at banks, credit unions, neobanks, fintechs, and payments companies—anyone responsible for launching new use cases or keeping fraud and AML programs current.
Can Vera Code help with regulatory changes?
Yes. Vera Code can compare new regulatory guidance, such as a FINTRAC bulletin, against an existing AML ruleset, propose changes that close coverage gaps, and implement them after approval. Thresholds and final decisions stay with the institution.
What are AI agents for fraud and AML?
AI agents for fraud and AML are AI systems that take action on behalf of fraud and compliance teams—analyzing data, building detection logic, configuring workflows, and drafting reports—based on natural-language instructions. Unlike chat assistants that only answer questions, they execute tasks inside the fraud and AML platform, ideally with human approval and full auditability.
See Vera Code in Action
Your expertise. Vera's execution. Your team in control. Vera Code gives fraud and AML teams a faster way to launch new programs and keep them ahead of evolving threats. Request a demo to see how Vera Code can help your team build and adapt complete fraud and AML programs in days.
