Startup Uses A.I. to Speed Up Financial ComplianceBy
Balderton Capital leads funding round for ComplyAdvantage
System scouts databases, reads news for signs of trouble
A U.K. startup that uses artificial intelligence to help banks and other financial firms with anti-money laundering compliance received $8.2 million to fund its expansion in Europe and North America.
The financing for London-based ComplyAdvantage is being lead by British venture capital firm Balderton Capital. The company, which said in a statement it has 200 clients globally, said it would use the money to expand its operations, adding to its current team of about 50 employees. It will also open a sales office in New York this week.
Complying with the growing body of anti-money laundering and know-your-customer regulations around the world is becoming an expensive headache for financial institutions. Consulting firm Accenture says the cost of complying with these rules has risen 50 percent in the past three years as a number of new regulations have come into force around the world. The U.K. Financial Conduct Authority commissioned a report by consulting firm John Howell in 2015 which found that at one large British bank, the cost of compliance per customer had risen from to 300 pounds ($366) per customer from about 65 pounds over the course of the previous two years.
"There are 7 billion people in the world and you are presented with this one individual, how do you know who is bad?" said Charles Delingpole, ComplyAdvantage’s chief executive officer and founder, in an interview with Bloomberg earlier this week. Delingpole previously co-founded London-based peer-to-peer business lender MarketInvoice Ltd.
ComplyAdvantage allows financial institutions to search prospective customer names through a proprietary database that includes lists of individuals sanctioned by entities ranging from the U.S. government to the United Nations and lists of known-criminals compiled by organizations like Interpol. It checks names against lists of politicians and government officials.
It also looks at data from regional regulatory authorities in several countries as well as information taken from press reports about crime. The system uses machine-learning to identify the names of people accused or convicted of crimes and weights what it uncovers according to the type of crime involved and whether the individual was actually convicted or merely accused or investigated.
The biggest problem with existing automated compliance systems, Delingpole said, is that they generate too many false positives -- finding individuals with similar names, for instance, even though they are not the person who needs to be assessed. He said that in trials with one client -- who he declined to name -- ComplyAdvantage was able to reduce these false positives by 60 percent.
Tim Bunting, the general partner at Balderton, said in a statement that ComplyAdvantage was unique because it was building its own database, not merely creating a new user interface to interact with existing ones. He said that he considered compliance "one of the few remaining large industries that is still ripe for digital disruption."
ComplyAdvantage prices its product according to the volume of searches a customer performs on the system. Customers can chose to simply flag possible issues with a potential customer -- to allow their own systems to quickly reach a yes or no decision -- or they can actually delve into the database to do a more detailed investigation.
Delingpole said he became interested in the challenges of compliance during his time at MarketInvoice and said he spent two years building ComplyAdvantage’s technology platform.
He said he initially trialed the system with payment firms that send remittances from the Somali immigrant community in the U.K. back to their homeland. "The problem we were wrestling with was how do you insure some of this money doesn’t go to Al-Shabaab," Delingpole said, referring to the Somalia-based militant group. "How do you make sure it goes to someone who genuinely needs the money?"
To continue reading this article you must be a Bloomberg Professional Service Subscriber.
If you believe that you may have received this message in error please let us know.