Blog

Recent data from the Capgemini Research Institute has found that AI agents could deliver up to $450 billion in economic value by 2028 – and financial institutions across the board are ‘actively’ moving key customer-facing processes to AI agents, according to a report released by the company earlier this month. Fraud detection, cited by 64% of respondents, is the second most common use case for deploying agents at scale, behind customer service (75%).

Super.com is deploying its own AI fraud agent to, in the words of Sarah Boehmer, senior director, payment processing and fraud strategy, not only detect fraud, but minimize friction for good customers. The company, best known originally for its online travel agency, has been steadily expanding into fintech -- it now offers a secured card, cash advances, and credit-building tools, among others.

With the help of a tech startup, the agent has been custom-built for the Super.com earnings app to review customers who trigger particular thresholds. “Our fraud and product teams never thought it would be possible to review 100% of a population, but with AI and the level of automation it allows, it is feasible” explains Boehmer.

Boehmer’s team includes more than a dozen specialists — from disputes analysts to senior managers — and works cross-functionally with product, engineering, and legal, among other teams. “Leading fraud in a growth-focused business requires the team to work proactively across the org, because fraud may not always be top-of-mind for others,” she notes, but adds: “It’s really exciting to be working in payments and fraud at a time when AI is just starting to take off, and we’re figuring out how to incorporate it into our processes.”

Developing the agent has been a boon for Boehmer’s team. She admits she was initially worried that some of her team, particularly the more junior members, may be concerned about AI’s impact. Yet wrestling with questions such as how much data to give the agent, and what the optimal QA strategy is, has helped. “It’s been a really motivating project for them to figure out how to continually improve it,” says Boehmer.

As Super.com has moved into fintech, the fraud team -- and Boehmer’s role -- has evolved with it. Having joined in 2021 as director of financial operations, she also took on payments at the start of 2024: “I saw a lot of opportunities for Super.com to level up and make payments more strategic,” says Boehmer. “It’s been really great for me to shift my career more in the direction of fintech, as the company has moved in that direction.”

Yet building the division from the ground up is not easy, particularly as an evolving business. “When the team first dug into fraud for the online travel agency business, it was mostly stolen credit cards, friendly fraud, and basic patterns,” says Boehmer. “As Super.com expanded into fintech, the fraud landscape -- and the team’s work -- became increasingly identity-focused.”

“What I’ve realised is – I don’t want to overgeneralise and say everyone’s identity is out there on the dark web to buy for any price, but there’s been so many data breaches and compromises, it’s so easy to get a hold of stolen identities and use them to break into fintechs,” Boehmer adds. “So people leading fraud programmes, like me, are having to think at the next level and come with a layered approach.”

This ranges from behavioural checks to device intelligence; again, an area where Super.com – and the fraud team – has had to learn fast. “It’s kind of a cliché, but it is like an arms race,” she notes. “Our fraud team is constantly adapting to keep up with increasingly sophisticated tactics.”

It is much the same story with AI; another arms race, as Boehmer explains. Take the example of a user applying for Super’s card, who will have to identify themselves with a government ID and a selfie. This causes its own problems, even with very advanced tools. “The fraud team has had to go well beyond face value – AI has enabled bad actors to mask their identities in far more sophisticated ways, which in turn has required us to adapt our strategy,” explains Boehmer.

“The flip side of that is those of us working in fraud prevention are also trying to use AI to keep up with them,” Boehmer adds. “We’re trying to use it more in ways to look beyond the surface, look behind the scenes, look for patterns that we may not have been able to see before.”

For building out agents at scale, Boehmer notes the importance of strong governance processes – again noting the question of how much data. “We could give [an agent] so much power,” she says. “Because it has so much access to data, we need to make sure we have security around that data. The more you have it do, the more control it has, and so you just need to have quality processes put around that.

“So the team is working to find the right balance of use cases that are good for customers, good for our team overall, and good for the business, while putting in the right level of governance and controls.”

Sarah Boehmer is speaking at Fintech Connect on December 2-3, in a session entitled ‘Tackling Cyber Threats in a New Era of Digital Payments.’ Find out more about the session and register your place today.