“It’s not enough to have a 1-800 number that people call, and the person still says it’ll take 90 days. It’s truly about that first touch resolution.”
Co-founder and CEO Shanthi Shanmugam’s vision for Casap is to become the FICO of first-party fraud scoring. Casap is quickly achieving that goal by viewing dispute resolution as a loyalty-building opportunity.
Casap recently announced a $25 million Series A round, bringing its total funding to $33.5 million, a reported record for venture investment in a payment disputes company. Emergence Capital led the Series A; additional participants included Lightspeed Venture Partners, Primary Venture Partners and SoFi.
“Casap is building the decision engine for post-transaction risk: scoring first-party fraud, automating exceptions, and streamlining ATO resolution. Wherever institutions depend on people navigating siloed systems, Casap provides the logic layer.” Lightspeed Ventures Partner Aaron Frank said in a LinkedIn post. “This is a $100B+ market defined by legacy systems and operational drag.”
In an interview with Fintech Nexus, Shanmugam said three factors are shaping the new reality of financial operations infrastructure. For one, consumer expectations are different than a decade ago. They expect an instantaneous, friction-free experience when moving money.
Next, first-party fraud is a growing problem. It now accounts for up to 50 per cent of fraud losses. Consumers are increasingly taking advantage of rules in their favor, and file disputes as a way of recouping money.
The third element is the technology that’s becoming ubiquitous in the financial services industry: AI, and where it can take the dispute process. Casap and others are betting AI can help institutions navigate the litany of rules governing consumer interactions.
“Generative AI is really good at taking unstructured text, making meaning of it, and communicating like a human,” Shanmugam said. “Now, the communication with the consumer and the merchant can be done at a very high level of quality, at a scale that wasn’t able to be done before. Combine that with traditional machine learning and all this data that you’re getting in one place, and now you have this score.
Casap designed AI agents that handle the full dispute lifecycle in one system. The platform analyzes evidence, predicts outcomes, and automates credit issuance, chargeback filing and merchant responses. The fraud score identifies risky customers and merchants to reduce disputes proactively.
Shanmugam’s career has been shaped by how technology can make positive differences in people’s lives. Following a stint at Meta, she worked on cryptocurrency investing as an early Robinhood employee. Like all financially-focused companies, Robinhood sought primacy with its customers.
Then came Gamestop. Shanmugam developed its 24/7 call and chat support services. She said a fiserv couldn’t foster that primacy if it couldn’t explain its actions.
“It’s not enough to have a 1-800 number that people call, and the person still says it’ll take 90 days,” Shanmugam explained. “It’s truly about that first touch resolution.
“What if we can go better than that, not even wait for them to call? We call and ask if they need help. That customer’s going to say, ‘Whoa, you get me.’”
Pair that with co-founder Saisi Peter’s product management experience with Chime, where they indicated negative dispute experiences were a top reason why customers left.
Peter and Shanmugam asked each other why it was so hard for institutions to figure out how to refund money. The culprits include data silos, a myriad of regulations, confusing network rules, and chargebacks outsourced to firms with no skin in the game. Increased skepticism from first-party fraud makes the problem even harder.
“We saw this recent problem of first-party fraud increasing since COVID,” Shanmugam said. “We just solved this problem on our own at these two different institutions…and we saw why someone chooses an institution is because of how they handle a dispute. ChatGPT just came out, and we thought there’s probably something really exciting we can do there.”
In the past, dispute resolution was a one-size-fits-all process, with processors paid to handle it in an indeterminate amount of time. Shanmugam said it’s a black box, with customers losing disputes they should win. Someone’s trust is lost in the process.
That trust can be maintained, and even increased, with meaningful friction, where customers see their issue being addressed. Say a consumer questions a strange Amazon address on their monthly statement. Casap’s system uses AI to scan past transactions to identify similar purchases or merchants and presents that to the consumer.
Was that friction? Shanmugam said it depends on your lens. Is it harder to file a dispute, or was it resolved faster than filing a dispute and waiting 30 days for a response? If a frequent disputer was transferred to a human for the last few questions, is that effective?
“You can find a way to balance the two, but it’s about being intentional when we think about the design,” Shanmugam said. “What are we trying to accomplish while keeping what at bay? What’s our target metric? What’s our category? Fast resolution without losing the shirts off our backs to fraud.”
Shanmugam said using this often-fraught experience to build loyalty begins with consumer empathy. They either didn’t make a transaction, or the merchant erred.
Regardless of the reason, how can the institution give them a 10-star experience? Maybe an immediate refund? Can the customer feel that their institution has their back?
The foundation is in determining honest customers, the ones with high Casap scores. That gives institutions confidence in winning disputes.
Casap’s results are strong, with many customers seeing positive ROI within weeks–leading for requests that the company help with other problems. Shanmugam said multi-player fraud investigations are one area where AI can harmonize multiple workflows. Payments and the myriad of firms and tools involved are another.
“When I asked dispute investigators what their favorite part about investigating disputes is, they say actually figuring out who’s telling the truth and not,” Shanmugam said. “When I asked them how much time they spend on that, they said maybe two% ;the other 98%is clicking on buttons, filing reports, and sending emails. That’s what we’re dealing with right now.”
Looking ahead, Shanmugam is watching how quickly GenAI compounds itself. Can LLMs like ChatGPT improve customer service?
“You’re having an honest, empathetic, trust-building conversation with someone at the institution that not only takes your information, but can also tell you what’s going to happen next and maybe resolve your case,” she said. “But lo and behold, it’s voice AI. Some of this AI support, like voice agents, is getting so good, they’re better than me; you could be talking to them.
As budget preparation season begins, Shanmugam said AI lets decision makers revisit tasks that in the past were addressed in one way, and it was a fait accompli. As they take a fresh look, Shanmugam wants them to think of Casap first.
“Having so much more money within a month of going live, and losing less money to fraud…who doesn’t want that, right?” she concluded.
