Introducing early access for Case Review Agents: AI decisioning for high-stakes identity decisions
Every day, your review team makes hundreds of decisions that determine who gets access to your platform. These decisions carry a lot of weight. Get them right, and you protect your business while delivering a seamless user experience. Get them wrong, and you either block legitimate users or open the door to fraud.
As your business scales, these decisions get harder to manage. Case volume climbs, fraud tactics shift, and regulatory expectations evolve. All the while, your review team is buried in a never-ending queue.
Some teams try to keep up by hiring more reviewers or outsourcing support to offshore providers, but consistency becomes harder to maintain at scale. Others layer in third-party AI decisioning tools that rely on API exports, which means key signals from the verification flow don’t carry through to the decisioning layer.
Working with enterprises across industries, the pain points in KYC and KYB reviews always boil down to the same question: how do you make reviews faster without sacrificing decision quality? This question prompted us to build Case Review Agents, our in-platform decisioning layer designed to help teams move faster while maintaining trust.
Better decisions start with better context
Most agentic case review tools promise to reduce prep work for analysts and accelerate case queues. At Persona, faster reviews are just the beginning.
What truly sets Persona’s Case Review Agents apart is the firsthand context they operate on, their ability to reflect how your team makes decisions, and hands-on implementation paired with rigorous backtesting led by Persona’s identity experts.
Built on the identity data organizations already have in Persona
Case Review Agents operate natively inside Persona, analyzing raw verification data, including the actual documents submitted, liveness checks performed, and risk signals captured during a flow. This gives them a level of context no outside tool can replicate. Our agents reason directly over your case data, and as your data grows, they get better at spotting emerging fraud patterns.
In contrast, third-party tools require data exports via API, meaning identity data leaves your platform and loses critical context, like raw verification artifacts and behavioral signals.
For example, a fraudster using a virtual camera to inject a deepfake might pass a liveness check that only analyzes the video. But inside Persona, risk signals like a perfectly still gyroscope, distraction events mid-flow, and camera metadata mismatches tell a different story. Those risk signals may not survive an API export.
Some companies also use customer data to improve their models, meaning your customers’ sensitive identity information could potentially be used to train a system that also serves your competitors. Persona only uses your data to improve your agents.
Calibrated to your judgment and your policies
Most AI review tools are configured to match written SOPs, then pushed to production. If backtesting happens, it typically just verifies that rules and thresholds behave as expected.
Persona’s agents can ingest SOPs as one input into how case review should work. But SOPs only capture what’s written in policy. Your organization’s historical case data reflects the judgment built through real decisions, such as weighing conflicting signals and handling edge cases. It’s harder to document, but crucial for calibrating agents to align with a team's actual decisions.
That's why Persona rigorously backtests against each organization’s historical data before deploying an agent. In practice, Case Review Agents consistently achieve 90%+ accuracy matching reviewers' past approve/decline decisions. Teams can also use the included dashboard to validate performance and improve the agent over time.
Implemented by identity experts who've seen it all
Every deployment is paired with hands-on implementation led by Persona's identity experts, who’ve supported hundreds of identity programs and millions of verification decisions.
They know which patterns actually matter for various use cases, which edge cases to stress-test, and how to tune agent behavior to different risk profiles. Persona also works closely with teams to get their agents’ decisions right before going live, and helps monitor and refine performance post-launch.
How Case Review Agents work
Case Review Agents combine a real-time review process, purpose-built decisioning features, and enterprise-grade controls to support the full case life cycle.
A real-time, end-to-end review process
The moment a case is ready, Case Review Agents get to work, reducing backlogs and time zone delays that human reviewers may face.
A case is created after a user completes an Inquiry or another event triggers a review.
The agent reasons over the full picture available in Persona (e.g, case data, account history, Inquiry results, and any external context pulled in through your integrations).
The agent makes a recommendation to approve, decline, or escalate with a structured summary of the reasoning. A complete audit trail logs every signal considered and conclusion reached.
Your team takes action, either automatically through the agent or with human review for complex cases.
Continuous improvement refines agent behavior by adjusting instructions, capabilities, and data access based on decision outcomes and your team’s feedback.
Features built for identity review
Case Review Agents come equipped with features designed to help teams move faster and make more accurate decisions.
Risk scoring: Agents assign risk scores based on your criteria and thresholds, enabling automatic handling of low-risk cases while directing reviewer attention to higher-risk ones.
Automated case documentation: Agents generate structured case summaries using available data to fill fields. This reduces the time analysts spend on documentation and minimizes AI hallucinations, since outputs are grounded in actual case data rather than inferences.
Fully connected workflows: Agents integrate with tools like Slack and Salesforce, enabling teams to receive updates, review cases, and take action directly within existing workflows. Integrations also allow agents to incorporate external context, like deal stage or CRM data, so decisions reflect the full customer picture.
Full case context and memory: Agents maintain memory of case context, eliminating the need for analysts to repeatedly re-provide information and enabling more consistent, informed decisions.
Built with privacy and security at the core
Identity data is some of the most sensitive information your organization handles. Case Review Agents are built with that in mind. Persona never shares your organization’s data with LLM providers for training, and Case Review Agents don’t share data across customers. Each customer's agent runs in its own isolated environment, and all conversations are encrypted.
You stay in full control of how your agent operates:
Model provider: Choose the LLM that fits your technical and compliance requirements.
Data retention: Determine how long data is stored and who can access it.
PII masking: Configure access controls and ensure agents and reviewers only see what they’re authorized to access.
Decision authority: Choose whether agents automate final decisions or surface recommendations for human review.
Access controls: Control data access and permissions across all your integrations.
Now in early access
Persona is launching an early access program for Case Review Agents. Ideal partners are struggling with manual review volume and are ready to define how AI-driven identity decisioning should work in practice.
Your organization’s users, fraud vectors, and workflows are all going AI-native. Persona is preparing for that world at every level, from the assistant helping you build your flows to the agents making decisions at scale.
Help us shape what comes next for our agents. Contact your account team to apply for early access.