One of the largest banks in Florida and the largest Chilean-owned bank in the U.S. said this week it is beginning to use artificial intelligence software to detect financial crime.
Bci Miami is one of the first banks in the U.S. to publicly acknowledge using AI this way, when many still consider the technology to be new, risky and unsanctioned by regulators.
But the bank is forging ahead after informing its U.S. regulators of the technology.
“Bci Miami is committed to serving our customers while working with regulators to continue to maintain a strong, compliant anti money laundering program,” said Michele Fernandez, its head of compliance. “We now have a more efficient risk management strategy in place.”
If you wanted to test AI for catching financial crime, you could find few places richer with opportunity than Miami, which has been dubbed "the organized crime capital of America."
There were 31,167 reports of identity theft in Florida in 2017, the second-highest state in the nation, according to statistics by the Federal Trade Commission. Those stolen identities are used in many different types of fraud.
"It seems that we are a hotbed for various types of fraud," said David Schwartz, president and CEO of the Florida International Bankers Association. "The things that make South Florida an attractive location for investors and people looking for a better place to live — our geography, weather, cultural diversity — attract good people and unfortunately, they also attract bad people."
As a result, banks in Miami are under special pressure to try to detect money laundering, human trafficking and other wrongdoing.
The bank is using three software modules from QuantaVerse. Other vendors of AI software that can be used to detect money laundering and other financial crime include Thetaray, Merlon Intelligence, ZestFinance, Ayasdi, Quantexa and IBM Watson.
One module by QuantaVerse is called Pre-TMS Entity Resolution & Risk Scoring, which is intended to reduce false-positive alerts by classifying the risk of each transacting party. (TMS stands for Transaction Monitoring System.) This software works to find missing information about people and companies and clean up information.
“It’s an analytics tool to understand the risks the entities are presenting to the bank,” said David McLaughlin, CEO and founder of QuantaVerse.
Handling a transaction for General Electric, for instance, would be considered low risk, especially for money laundering. A transaction conducted on behalf of a company whose owner was indicted for financial crimes in a previous job would be labeled high risk.
The second piece of software is called Alert Investigator. It helps human compliance officers investigate financial crime alerts.
“When a human gets an alert from a transaction monitoring system, they have to pull data from different files, search open source data, and put together an investigative case that allows someone to determine whether this needs to be reported as suspicious activity,” McLaughlin said. “We can automate 70% of that investigation. We can provide the investigator an automated financial crime report with natural language generation that includes a recommendation of whether to file a suspicious activity report or not or if further investigation is recommended."
The third component is a false-negative identifier. This looks at all the transactions that were not alerted and analyzing them to make sure nothing nefarious was missed. The bank’s top goal is to improve efficiency in its investigative process as it grows. Its second goal is to catch every instance of financial crime in its organization and prevent money launderers, drug traffickers and human traffickers from using its rails.
These tools come in to play together in detecting suspicious behavior in South Florida.
FinCEN has targeted certain types of businesses in specific Miami ZIP codes, looking for signs of illegal electronics exports and other trade-based money laundering schemes. Such orders force local banks to comb their databases to see if they serve any of the affected businesses.
Another complication for South Florida banks when it comes to detecting money laundering is that in the Latin American countries with which they tend to business, there are a lot of people with similar names. Distinguishing bad actors from innocent people is that much harder.
"You can get an incredible volumes of false positives you have to analyze, so being able to put all that data into a machine and have that machine be able to compile it all, review it all, analyze it, pull it from various sources can definitely go a long way to enhancing banks' programs," Schwartz said.
Miami banks spend a lot of resources, time, personnel, and money on compliance, he said.
"If you talk to the local regulators, they will tell you the banks in Miami have done a good job, not just in terms of their compliance program, but in cooperation with regulators locally," Schwartz said.
Schwartz knows of several banks locally and nationwide that are testing AI for anti-money- laundering purposes.
"There’s such a large volume of activity and the incidence of false positives is so high that you’re throwing a lot of resources at trying to analyze all these false positives,” he said. “And banks are spending a lot of time and money just analyzing and eliminating them. An AI engine can analyze in seconds a client's history, profile and the types of activities he's had going through that account. It would be easy and quick for the machine to understand that client's profile and match it against that activity. If there is an alert at the end of the day, it's still coming into human hands to review it, but at least the whole process of getting there will help reduce the number of false positives."
Editor at Large Penny Crosman welcomes feedback at firstname.lastname@example.org.