How Casap is rethinking dispute resolution with agentic AI

Payment disputes can take up to 90 days to resolve. Casap, an AI-native dispute and fraud platform, sees an opportunity to compress that timeline with agentic AI. Its AI agents work across the dispute lifecycle, handling tasks from claim intake and evidence gathering to fraud detection, chargeback filing, and customer communications.

The result is faster resolution, with institutions using the platform cutting resolution times from months to days while maintaining compliance and an auditable record of every decision. At Chartway Federal Credit Union, average resolution times fell from up to 3 months to 12 days, while per-claim costs dropped nearly 90%, write-offs declined 72%, and chargeback win rates improved by 95%. As a result of implementing Casap, the credit union generated $875,000 in first-year savings.

For this work, Casap was named AI Company of the Year at Tearsheet’s AI Innovation Awards 2026. Tearsheet spoke with Shanthi Shanmugam, CEO of Casap, to explore how the agentic AI-powered dispute platform works, where human judgment remains essential, and where agentic AI is headed in dispute management.

Q: How does an AI-native foundation change what Casap can do versus incumbents retrofitting AI onto legacy systems?

Shanthi Shanmugam, Casap: We built Casap with agentic AI from the ground up, including the first publicly available dispute model and API. Casap isn’t a workflow tool that routes tasks between people. It works more like an investigator who already has everything in front of them.

Dispute data has always been scattered across systems, which is a big part of why this work is so hard on the teams doing it. We bring it into one place. That’s what makes it possible to see patterns of first-party fraud that legacy infrastructure with an AI layer bolted on simply can’t see, and to score that risk before a provisional credit goes out the door.

Being AI-native also means compliance logic is foundational rather than layered on top. Reg E, Reg Z, NACHA, and card network rules are all embedded in the workflow. That’s why automation and auditability grow together for us, rather than pulling against each other.​

Q: What signals tell Casap’s AI agents when they can act on their own and when a case needs human judgment?

Shanthi Shanmugam, Casap: Two signals: a predictive win score that gauges the strength of the case and our first-party score. Both come from models built on proprietary, institution-specific dispute data to determine confidence in a given case.

When confidence is high, and the dispute is clear-cut and low risk, the platform supports an instant provisional credit decision, which is exactly what someone waiting on their money deserves. When a case is more complex or carries higher risk, human review is recommended by design. If the platform sees a serial bad-faith disputer, it flags that to the investigator on the case, because those calls need real judgment and a person’s read on the situation. With our managed service, you can outsource the investigation to us.

The models are there to guide case handling and decision-making, not to take it over.

Q: How do you make an AI agent’s reasoning defensible when a financial institution needs to explain a decision months later?

Shanthi Shanmugam, Casap: Every dispute lives in a centralized case record that tracks required actions, supporting documentation, and regulatory deadlines from intake through resolution. That gives you a full audit trail for every automated decision. And because predictive win scores guide rather than replace human decision-making on complex cases, there’s always a documented rationale tied to the record. We also monitor model performance for accuracy and bias on an ongoing basis so that institutions can point to the validation process behind a decision, not just the decision itself.

Say an examiner comes back to a claim from nine months ago. The compliance team pulls up the case and reads it start to finish. How the claim came in, what the agent found in the account and transaction history, why it scored and routed the way it did, when provisional credit went out against the Reg E deadline, and what the consumer was told at the end. The platform tracks deadlines, and dispute data can be exported for regulatory reporting into the systems the compliance team already uses.

Q: Where do you see the biggest untapped opportunity for AI across the dispute lifecycle today?

Shanthi Shanmugam, Casap: Most AI investment has gone toward making the decision faster. Our experience is that the front end matters just as much, and it’s the part people tend to overlook: the very first interaction with the customer.

Casap’s intake is adaptive and conversational, and it does more than collect information. It assesses claim validity in real time and often resolves the situation by pointing the consumer to the merchant before a formal dispute is ever filed. Everyone comes out ahead when that happens.

The consumer-facing transparency layer is the other overlooked opportunity. Our customers love the dispute tracker, and it’s more than a nice service touch. Poor dispute experience and opaque timelines are the second-leading reason customers leave their financial institution. Keeping someone in the loop is a retention lever, not just a courtesy.

Q: How do you see agentic AI reshaping dispute management over the next three years, and where does Casap fit into that evolution?

Shanthi Shanmugam, Casap: Dispute management will stop being reactive. Today, it’s case-by-case, and it starts at the point of dispute. Over the next few years, we expect it to become a proactive risk intelligence function, where fraud scoring happens earlier and runs continuously.

AI decisioning is already the baseline expectation in other industries. Our job is to keep proving that regulated, high-stakes financial workflows can be automated end-to-end without sacrificing auditability, and that fraud losses can be a fraction of what they are today. Institutions on Casap are already cutting fraud losses by half.

That’s what excites me most, because it opens the door for smaller institutions to access capabilities that were once reserved for the largest, best-resourced banks.