Even for experienced bankers schooled in compliance guidelines and use of legal software, decoding financial regulations is no easy task.

“Historically, governance, risk and compliance tools have not been great at that first line of defense,” said Michael Curry, vice president of engineering at IBM's Watson Financial Services. “They're hard to understand, they require a lot of training and to be able to train that many people to really be able to understand the tools and be able to use them has been too difficult of an exercise for most financial institutions.”

To update its governance risk and compliance products, the tech giant has sought to blend the data analytics software of Armanta, which it acquired in May, with both the natural language processing skills of Watson and the regulatory expertise of Promontory Financial Group, another IBM subsidiary.

IBM said the new regtech combination has already yielded better results.

For example, the technology was able to identify when a bank’s alert system was falsely flagging “scuba” within a company’s financial information because it contained the word “Cuba,” Curry said.

IBM said Watson was able to identify 30% to 50% of bank anti-money-laundering alerts as false positives, and reduce the time it took for a banker to perform customer due diligence on a standard business from 13.5 minutes to five minutes and 20 seconds.

Michael Curry, vice president of engineering at IBM's Watson Financial Services.
Michael Curry, vice president of engineering at IBM's Watson Financial Services.

Curry’s team studied the design of OpenPages 8.0 and IBM Financial Crimes, which both facilitate government risk and compliance, by testing both products with front line bank employees and redesigning areas where the bankers got stuck.

The new combination utilizes libraries of regulatory requirements, allowing users to match specific rules with issues in their compliance program to ensure proper controls. Armanta will also provide employees with what Curry calls an “in-memory cube” to help bankers see risk issues across multiple departments.

“The idea behind this is very similar to a pivot table in Excel, where you analyze and aggregate data from many different dimensions,” Curry said. “In the case of what we do, you may want to roll up risk based on department, portfolio, currency, country, product or any number of other dimensions. The ability to do this level of aggregation and calculation across any number of dimensions in real time is unique to our technology.”

When IBM partnered with Armanta before acquiring it, IBM used the analytics company mainly to assess financial risk for its clients.

“It’s more natural for Armanta to address security and fraud as opposed to financial risk,” said Ryan Gilbert, a partner at Propel Venture Partners. “Those assessments are a lot more fact-, rules- and knowledge-based. Assessing financial risk involves a very different set of processes and considerations and is seldom clear cut.”

Outside of improving efficiency and cutting cost, the most important question that IBM has had to answer since its entrance into regtech in June 2017 is whether or not banks can trust their regtech software.

“When a bank hires a big-name vendor like IBM, part of hiring that vendor is so they can say to regulators, ‘Hey we spent the money on this vendor so we don’t have to worry about that,’ ” said Danielle Tierney, senior analyst at Aite Group. “It’s different if you hire a no-name vendor where there would be more oversight required.”

Still, for natural processing systems like Watson, the greatest hurdle to reliable regtech services is a language barrier.

“There is a lot of variability in how regulations are written in Germany, versus Singapore versus the U.K. versus the U.S.,” Curry said. “The actual language is different. But also the wording type, the way clauses are written, they’re all different… There’s a lot of room for improvement around how regulations are written.”

Even with new technology, industry experts expect the compliance process to stay human for a while.

“I don’t believe that we’ll ever achieve 100% robo in financial services until regulators themselves become 100% robo,” Gilbert said.