AI — as software did before it — is eating the world, and pretty much every industry is trying to work out how to get the most out of the technology. Fintech, which has created some of Europe’s most valuable tech companies to date, is particularly keen to cash in.
Swedish BNPL scaleup Klarna is AI cheerleader in chief, recently announcing that it’s made big cost reductions in both customer service and marketing, thanks to the technology.
But use cases like these, which could be implemented in a large number of industries, are just the tip of the iceberg.
Fintechs are also using AI for specialised applications — to try and make our money more secure, and the way we manage it more efficient — as companies around the world battle to stay ahead of the pack.
Fraud
One of the most well-established ways that fintechs have been using AIis to help address regulatory responsibilities like Know Your Customer (KYC) and Anti Money Laundering (AML) — compliance standards to reduce fraud, corruption and money laundering.
“I probably first saw this in the compliance space three or four years ago,” says Michael Kent, venture partner at VC firm Headline and serial fintech founder.
“Ultimately what we're trying to make sure of in most of the compliance stuff is that the person is who they say they are, and there's no risk associated with that person.”
Kent points to London–based regtech ComplyAdvantage, founded in 2014, which has developed a product to help fintechs know if they are potentially dealing with high-risk customers, using AI to scan databases of “watchlists, politically exposed persons and adverse media.” Newer entrants working in the space include Prague-based Resistant AI, which helps its users protect against document fraud, and London-based Tunic Pay, which uses AI to fight against payment-based scams.
Research from Evident Insights suggests that compliance is also the biggest area of AI investment for legacy financial institutions.
But some startups are starting to push the boundaries even further.
Kevin Levitt, head of financial services at US chip making giant Nvidia where he helps fintechs implement AI, says that one of its clients — Dutch digital bank Bunq — is using the technology to create synthetic data (an automated technique for generating artificial information) to identify novel fraud techniques that haven’t even been dreamt up by criminals.
“When you're fighting fraud, you're looking for anomalies in the data. And sometimes there aren't enough of them, because there are new ways of attacking financial institutions and consumers,” he tells Sifted.
“You can use generative AI to produce synthetic data that actually represents new innovative fraud techniques and you can train the model to then recognise those so that you can fight them before they're even developed by the bad guys.”
Other use cases
AI is being tested out on plenty of other use cases, too. Paris-headquartered lending platform Defacto uses AI to help make decisions on the credit-worthiness of users — and has a slim-line team as a result, says Kent.
“That business is currently doing the best part of half a billion euros worth of lending with about 20 people. That just wouldn't be possible historically,” he says. “It's not the sole arbiter of truth, and you probably wouldn't want it to be for credit decisioning, but they're definitely leveraging it.”
Levitt says that Nvidia has also been exploring how fintechs might be able to use generative AI to smooth ID verification processes and improve user experience for customers.
“We've also been talking to companies this week about how you can use generative AI to actually generate credentials for transactions before somebody even gets to the transaction page of a website,” he explains. “So, how can you anticipate a consumer’s intentions and go ahead and permission them for a certain transaction before they even get into that workflow?”
Another novel AI use case: Bunq is using language models to power its new, in-app personal assistant that helps customers manage their money, rather than just responding to simple customer service requests.
“It’s about moving from a scripted chatbot that can only handle a certain number of inquiries, to a generative AI capability that can service far more questions and provide accurate answers to help consumers on their financial journey,” says Levitt. “So it's very much financial empowerment, financial enablement for their customers.”
Caution
But some warn that applying generative AI to sensitive areas that cross over into financial advice could be risky, given the technology’s unreliability and tendency to make mistakes.
And as some fintechs rush to adopt AI to stay ahead, Alexandra Rivas-Gale, VP of product at London-based digital bank and payments platform ClearBank, says that her company’s strategy is to go a bit more slowly and deliberately. It’s primarily using the tech to make operational processes more efficient, for now.
“I say, ‘quality of AI over quantity of AI,’” she tells Sifted. “AI is only as good as the information that it’s being fed. Obviously, we need to make sure there's the right controls and guardrails around that.”
