UBS Taps Machine Learning for KYC

UBS is banking on machine learning to solve its know-your-customer (KYC) problem and is undertaking a project to use the technology to match information and find anomalies in customer information.

The Swiss bank is working with unnamed partners to build a KYC and anti-money laundering (AML) platform that uses machine learning, particularly natural-language processing (NLP), to take in data from public sources and connect it to customer information.

Mike Dargan, group chief information officer at UBS, says the bank is constantly looking at projects that add more value to the company, including KYC and AML activities. UBS is currently working with partners to bring more machine learning and natural-language processing to KYC.

“We’re working with a few partners on this—I actually prefer not to disclose names, but we are already working with them,” Dargan says. “It’s something that’s ongoing; though with artificial intelligence (AI) I don’t think you’re ever done-done. Ideally, you want to start early-ish in terms of what you want to do. It does need training to be better so we are ingesting the data forms and then seeing the outcomes and then comparing that to what we do already to get a better solution.”

He says the project will eventually be rolled out to the whole bank and cover many of its activities while its partners are looking to offer the KYC platform to other banks, possibly as a utility.

UBS’s machine learning KYC project will take in data about their customers from publicly available sources like disclosures or newswire services. Once the information has been digested, Dargan says it will match the information to each entity and look for any anomalies within the data that needs to be corrected. The platform aims to provide better information on customers that can be regularly updated.  

Dargan says UBS has been working with its partners on the KYC project for about a year now. 

KYC has been an operational problem because it is an integral but non-revenue generating activity for many financial institutions. These firms not only have to know their customers because of the law, but it is also important so they can better provide services to them and not double up on offerings. The industry has tried different ways of addressing KYC including the use of utilities; however, KYC utilities have not yet reached critical mass and therefore many of the purported benefits are unrealized.

Recent KYC projects include Societe Generale’s KYC offering via API on its single-dealer SG Markets platform, Refinitiv, Trulioo, Qual-ID’s tool meant to combat deepfakes, and Swift’s new API to automate its KYC Registry.

For UBS, KYC and AML offer opportunities to explore using technologies like AI. Dargan specifically points out machine learning as the best type of AI to use with KYC especially now that the technology can make connections between two entries of the same person but with typos.

According to Dargan, machine-learning technology is a good fit for activities like KYC, which provides a lot of raw data.

“If you think about some of these areas—KYC or AML or stuff like that—ultimately it’s a data problem. KYC or AML are areas where you’re looking for the anomaly in the data,” Dargan says. “You’re looking for those spikes in activity form AML—same in KYC—so you should be able to ingest data, whether or not that is unstructured data, then use NLP to process that data. Depending on the client, whether it’s an individual or corporate or an institution, you should be able to ingest that from public sources.”