Ep. 15: How can AI be used in fraud detection? Riya Jagetia of SocratiX AI explains AI co-workers for fraud teams
EPISODE GUESTS
Riya Jagetia is the Co-Founder and CEO of SocratiX AI, where she is building AI-driven technology for fraud and risk teams. Before founding SocratiX, Riya spent four years working in fraud, including roles at DoorDash and Unit21 focused on fraud detection, product development and AI systems.
Her current focus is reducing the manual investigative work carried out by fraud teams and using AI agents to work alongside analysts within their existing systems.
SHOW NOTES
Key Topics Discussed:
How AI can reduce manual fraud reviews and investigative work
What AI co-workers mean for fraud operations
Moving beyond fraud dashboards towards AI agents that take action
How AI agents can learn from analyst behaviour and new data
Deploying AI without replacing existing fraud technology
Human-in-the-loop controls for higher-risk fraud workflows
QA and oversight when AI agents begin taking automated actions
Why fraud, risk, compliance and AML should be treated as separate disciplines
Episode Summary:
This episode gets into how AI can be used in fraud detection and fraud operations, and why it matters for financial institutions and fintech teams dealing with large amounts of manual investigative work.
Riya Jagetia, Co-Founder and CEO of SocratiX AI, joins Justin Hanna at Money20/20 USA to discuss AI co-workers for fraud teams, how agents can work alongside existing systems and where human oversight still matters.
The conversation looks at a simple problem facing fraud teams: adding another dashboard does not necessarily reduce the work. Analysts can still be left checking information across multiple systems, reviewing cases and manually deciding what action needs to happen next.
Riya's view is that AI becomes much more useful when it moves beyond decision support and starts helping to complete the work itself.
How can AI reduce manual fraud reviews?
For Riya, the opportunity is to use AI agents more like assistants or co-workers.
Rather than giving fraud analysts another tool they need to monitor, an AI agent can bring together information from different sources, generate insights and complete parts of an investigation or workflow on the analyst's behalf.
That distinction matters. Fraud operations often involve repetitive manual reviews, and simply adding more data or another dashboard does not remove that workload.
SocratiX's approach is designed around agents that can observe how analysts work and use that input to help complete similar workflows.
The agents can also adapt as new information becomes available. Riya gives the example of adding a new data vendor. With a traditional system, a team may have to update dashboards, logic and rules manually. An AI agent can instead learn how analysts use that new information and reflect it in the workflow.
Do AI fraud tools need to replace existing systems?
Not necessarily.
One of the biggest barriers to introducing new technology within banks and fintech businesses is integration. Large technology deployments can take long enough that the original product or requirement has already moved on by the time everything is live.
Riya argues that AI fraud technology can take a different route.
Rather than asking customers to transfer everything onto an entirely new platform, SocratiX connects to the systems teams already use. Its agents then operate across those tools in a way intended to mirror how a person would complete the workflow.
Riya says SocratiX's first customer was live within three days and that most customers can be live within a week.
For fraud leaders, the commercial point is significant. Faster implementation gives teams an opportunity to test whether the technology genuinely improves operations before making a much bigger long-term commitment.
Where does human-in-the-loop AI fit?
Giving an AI agent the ability to act naturally raises questions around risk and control.
Riya describes a staged approach. When customers first deploy SocratiX, they can have a human review every AI agent output before an action is taken. Once teams have seen the results and built confidence in the system, more of those actions can be automated.
Human oversight does not disappear at that point. Riya recommends continuing QA and QC in much the same way a business would review the work of a human analyst. A proportion of AI-generated reviews can be checked each week to make sure the agent is still responding appropriately as fraud patterns, data and the organisation's risk appetite change.
This is where human-in-the-loop AI has practical value. The objective is not simply to automate as much as possible. Teams need controls that allow automation to increase without losing visibility over how decisions are being made.
Why fraud, risk and compliance should not be lumped together
Riya's Shelf of Shame nomination is FRAML, the term commonly used to bring fraud and anti-money laundering together.
Her issue is less with the technology and more with what happens when businesses assume fraud, risk and compliance teams all have the same needs.
They do not. Riya argues that these functions have separate workflows, motivations and priorities. They may use some of the same software, but product teams still need to understand what each group is trying to achieve.
For vendors serving financial institutions, that is an important product lesson. Combining several teams into one broad category can make the proposition easier to describe, but it can also hide the operational differences that determine whether the software is actually useful.
FAQ
How can AI be used in fraud detection?
AI can support fraud teams by reviewing cases, bringing information together from multiple systems, identifying useful insights and completing parts of investigative workflows. In the approach Riya discusses, AI agents work alongside analysts rather than simply providing another dashboard for them to monitor.
How is AI used in fraud detection?
The episode focuses on AI agents supporting day-to-day fraud operations. They can learn from analyst input, use information from existing systems and help automate manual investigative tasks, while human reviewers remain involved where additional oversight is required.
The big takeaway: AI becomes more valuable to fraud teams when it can remove operational work rather than simply produce more information. For banks, fintechs and fraud leaders, that means judging AI on how well it fits existing workflows, how safely it can take action and whether it genuinely reduces manual effort. Get that right, and teams can automate more without losing control. Get it wrong, and AI risks becoming another system analysts have to manage.
MEET THE HOSTS
Co-Host and Co-Founder of The Payments Shed Podcast
Grant Evans
Grant Evans is a leading voice in the fintech industry and the creator of the widely followed ‘The Payments Shed Newsletter’. With more than 15 years experience shaping commercial strategy and driving partnership growth, he is recognised for turning complex topics such as embedded payments, BNPL, unified commerce, and open banking into clear, actionable insights that resonate with global audiences. Named a LinkedIn Top Voice in both 2024 and 2025, Grant has built a community of over 27,000 engaged professionals, merchants, and innovators who look to him for commentary on the trends redefining global commerce. A sought-after speaker and panelist, his thought leadership is regularly featured in financial services publications and at flagship industry events including Money 20/20, FTT Fintech and the Global RegTech Summit.
Co-Host and Co-Founder of The Payments Shed Podcast
Justin Hanna
Justin Hanna was recently named the #1 Head of Sales Top Voice by the National Sales Conference for good reason: he’s redefining what sales leadership looks like in the modern era. With deep B2B sales experience and a people-first approach, Justin earns trust through insight and practical strategy, not tired tactics. A respected voice in payments, he’s also built a 22,000-strong LinkedIn following by making complex topics relatable and actionable. His influence has been recognised widely: a LinkedIn Top Payment Systems Voice (2024), one of the top 30 voices shaping the future of payments, banking, and fintech (2025), and celebrated by the National Sales Conference as the #1 Head of Sales Top Voice. Known for challenging the status quo, Justin’s unfiltered take on leadership, culture, and growth resonates because it’s honest, and his ability to lead with both expertise and empathy has made him one of the most influential sales voices today.
EPISODE TRANSCRIPT
Justin Hanna: Welcome back to The Payments Shed Podcast. Today we are joined by Riya Jagetia, CEO and Co-Founder of SocratiX AI. Tell us a bit more about yourself.
Riya Jagetia: My name is Riya. I am the Co-Founder and CEO of SocratiX.
We started about six months ago, but I’ve previously been in the fraud space for the past four years, building solutions at DoorDash and before that on the vendor side at Unit21.
I realised that fraud teams deal with a lot of manual reviews and manual workflows, and wanted to build an AI solution that helps lighten the load of what those teams have to deal with.
Justin Hanna: Fantastic. The industry, as we know here at Money20/20, is full of tools that use AI to assist teams. What does AI mean to you when it comes to, instead of creating dashboard after dashboard, really helping businesses not have to go through all that manual labour?
Riya Jagetia: That’s a great question.
I think that’s why we really frame our solution as AI co-workers, because we don’t just want another tool that people have to take a look at, get information from and then action.
We really want to build something that can work next to our customers, action for them, bring them new insights and stitch together information from different sources, really acting like an assistant rather than just another tool that you have to check every day.
Justin Hanna: I think we see it quite often. Businesses, fintechs and banks roll out what they think is a new product internally. It takes two years for them to really roll it out, and by the time it’s live, it’s already behind the times.
Where do you think you guys with AI, I’m going to assume that it’s updating all the time and always getting more intelligent?
Riya Jagetia: That’s a great point. I think that’s one of the ways that our solution is different from traditional dashboards or tools.
It’s constantly evolving, constantly changing and updating based on human input, but also new data.
Suppose you bring in a new vendor and that vendor brings you more data that’s useful. You would usually have to update a dashboard, update your logic and your rules to make sure it’s taking that into account.
Our AI agents can just learn from what your analysts are doing to make sure it’s acting the same way.
Justin Hanna: Makes complete sense.
When we go and speak to most businesses, when trying to put new products and roll out new products across their teams, the issue we normally have is the challenges that come with it and the dread of having to roll out a brand new product.
What are your thoughts on the way that AI is going to roll out compared to how it is maybe being rolled out in fintech or banks today?
Riya Jagetia: That’s a great question.
I think that was one of the things we wanted to make sure was a major differentiator of what we’re building versus what potentially legacy customers are asking banks and fintechs to do.
We don’t say, send us all of your data onto our platform, do a really long integration and then we have all the bells and whistles on top.
Instead, we plug into customers’ existing systems.
We have a proprietary infrastructure that we’ve built that allows us to get customers up and running in a matter of days.
We have custom connectors and then we just work with their existing tools to complete workflows the way that a person would.
Justin Hanna: You mentioned prior to us recording how quick it is for businesses to get live and running with you.
I’m assuming that’s going to be a big differentiator for you as a business, because you’re doing a lot of the heavy lifting on their behalf.
Riya Jagetia: Exactly. It’s a huge differentiator because, like I said, our first customer was live within three days.
Most of our customers are live within a week at most, and that makes a big difference because they can test out the solution and look at the results before even entering into a long-term agreement.
That makes a world of difference.
Justin Hanna: I bet.
So, we talk about AI investigating fraud and I think here at Money20/20 I’ve probably spoken to three or four businesses that are changing the world when it comes to AI.
It’s risky, and we all know that.
How do you think you provide comfort to the risk teams and maybe people that are seeing the glass as half empty as opposed to half full?
Riya Jagetia: There are basically two things that we do.
The first is we allow them to have human-in-the-loop controls for any high-risk workflows. They could do it for all of their workflows.
What customers usually do is, for the first couple of weeks, they’ll have humans review every AI agent output before it takes an action.
Then they usually say, can we please have the AI agents automatically start to take action because the accuracy is really high?
At that point, we say yes, we can definitely take action automatically for you.
But just like you would for a person, we want you to have QA and QC.
So maybe 10% of the AI agent reviews every week, someone else takes a second look at, just to make sure that as your patterns change or as your risk appetite changes, our AI agent is reacting appropriately.
Justin Hanna: Makes sense. Thank you very much.
What are your thoughts so far on Money20/20? We know it’s an incredible event. 13,000 people. It’s a lot to take in. How have you found it so far?
Riya Jagetia: It’s been amazing.
I feel like AI is the name of the game everywhere. AI and probably stablecoins.
I think what’s been really interesting is trying to dig a step deeper or a level deeper with customers to figure out what they actually mean when they say AI.
How is it actually being used? What are the measurable results?
I think ROI is so important when you’re deploying new solutions.
It’s been fun to see all the information that customers here, or companies here, have to offer.
Justin Hanna: Fantastic.
Last thing here on the Shelf of Shame. We love it here at The Payments Shed Podcast, talking about what really grinds your gears.
Tell us a little bit more about what really gets to you here in the payments and fintech industry.
Riya Jagetia: I’m really in the fraud space, the fraud and compliance space, and there’s this term that was coined maybe five or six years ago called FRAML. It’s fraud and AML.
That really grinds my gears because I think it’s so important when you’re developing products to have a really clear definition of what every team does, what they care about and what their motivations are.
Fraud, risk and compliance actually are three completely separate workflows with separate motivations.
So it really grinds my gears when people try to combine them all into one because then you’re just not understanding the customer super well.
If I could change the world, I would say let’s start talking about these three teams separately, even if they use the same software.
That way we deeply understand their needs, their concerns and what to build for them.
Justin Hanna: Fantastic. So we need to retire the word FRAML.
Riya Jagetia: Retire that.
Just say fraud, risk, compliance or AML, and be really clear about what you’re solving for them, why and what their return is.
Justin Hanna: Fantastic. Thank you so much for your time. Enjoy the rest of the show.
Riya Jagetia: Thank you so much.
Join the podcast for a relaxed conversation where you can share your experience and perspective with people working across the payments industry.
Be a Guest on the Podcast
Sponsor the podcast and get your brand featured in front of the top players in the Payments and fintech industry.