WASHINGTON — IBM is set to announce new tools on Wednesday aimed at reducing the compliance burden of financial institutions for combating money laundering, and a host of other regulatory requirements and sifting through the mounds of data they collect by using Watson, its cognitive computer.
Watson has absorbed regulations from 200 different sources and been briefed on 60,000 regulatory citations as well as trained by former regulators who work for Promontory Financial Group, a subsidiary of IBM.
“We're announcing … the first wave of cognitive tools to assist in regulatory compliance,” said Eugene Ludwig, the founder and chief executive officer of Promontory. "These tools have been and will constantly be carefully calibrated so that they will meet regulatory expectations.”
The move is a bold step forward for AI-like solutions to regulatory compliance (IBM does not like to call Watson artificial intelligence, viewing it as different). It is designed to help provide financial institutions with a computer consultant that can help in myriad ways.
There are three distinct tools. One is directed at financial crime and is designed to help identify customers, flag potentially suspicious transactions and examine internal communications for possible fraud or improper activity. Another is aimed at general compliance with rules and regulations, including ensuring financial institutions meet regulatory standards. The third is designed to help banks handle the growing volumes of data held in their systems as they scale up, specifically to stay up-to-date on trading book regulations, liquidity analysis and valuation adjustment measures.
IBM stressed that in addition to helping banks manage and comply with their regulatory requirements, the programs are meant to be examiner-friendly.
“Given that all of these [tools] touch highly regulated areas, we believe this is something that has to be done with a great deal of care,” Ludwig said.
To ensure examiners are comfortable with it, IBM'S new tools are not intended to be a "black box," where AI spits out an answer without detailing the reasoning behind it. Instead, They are designed to help financial institutions justify their decision-making process with a clear paper trail.
“Examiners like to examine what went on,” Ludwig said. “So that's important, that the process and the data that go into a decision can be looked at.”
Alistair Rennie, the general manager for industry solutions at IBM, said “it’s more than just a log."
“It is presenting a hypothesis with evidence and a confidence score," he said.
Watson will also learn as it goes. “As experts in the bank make this final decision, the model gets better over time," Rennie said.
For example, with its anti-money-laundering tool, Watson will rely on external databases and publicly available information, in addition to transaction reports and other data held by financial institutions, to track suspicious activities more effectively, Promontory said. It will flag suspicious activity to the appropriate compliance officer, who then will make a decision about whether to file a report.
Additionally, by going beyond classical rules- and lexicon-based algorithms, Watson is designed to be able to identify suspicious behavior that is occurring inside the financial institution, according to Promontory.
Watson has already found that before committing a misdeed, U.S. traders tend to stop swearing while U.K. traders are more likely to dial up their use of profanity.
Financial institutions will be able to use Promontory’s new tools either by deploying them onto their IT infrastructure or by placing their data in IBM’s cloud.
“Some of this data is going to be sensitive data,” Rennie said. “The information has to apply to the highest standards of security.”
But he stressed that Watson would not use aggregate data from other firms in the industry to fine-tune its algorithms for individual companies.
“We are very clear from our relationships with clients that their data is their data,” Rennie said. “There is an absolute hardwall over that.”
If, however, financial institutions decide to come together to pool their information on matters like money laundering, IBM could facilitate that.
“It is certainly true that over time, particularly in the AML space, institutions may themselves want to come together and form a consortium,” Ludwig said, “again, [while] respecting the privacy of individual customers.”