EXCLUSIVE— For machine learning and other technologies to make regtech the future, firms are going to have to ditch their focus on keywords and build solutions that focus on cognitive technologies and behavior.

This is especially true when it comes to financial compliance requirements for banks, traders, or others in financial services, who need to keep track of (and stay compliant with) an increasingly large regulatory load.

“It’s all very much rules-based, it looks for a pattern, for keywords, so you end up with a lot of false positives,” Marc Andrews, vice president, Watson financial services for IBM, told Bank Innovation. Andrews referenced the infamous Libor case, where some regulators were searching for misconduct by highlighting the term “USD.”

Instead, regtech needs to highlight potential misconduct in real-time, according to Andrews, meaning solutions have to watch for a shift in behavior amongst traders, for instance, or by watching for regulatory changes that could impact a company.

With its move nearly two years ago into financial compliance, IBM’s Watson is tackling these problems with three solutions. One keeps track of regulatory changes, which is live and integrated with IBM’s OpenPages platform, used by institutions like HSBC, Fidelity, Morgan Stanley, and Westpac.

The other two involve conduct surveillance, mainly for traders, and the third deals with financial crimes such as money laundering.

“The characteristics of cognitive — being able to understand natural language, being able to understand ambiguity — that kind of learning aspect tied it perfectly to regulation and financial compliance,” said Andrews regarding the decision to direct Watson towards this area.

Now, nearly two years after this decision, IBM is focusing on enhancing and updating its three solutions for the new age by acquiring and integrating the technology of acquisitions like compliance consulting firm Promontory.

These upgrades will include expanding Watson’s conduct surveillance solutions for other areas of finance (like certain sales practices that could be observed at certain banks, for instance) as well as new solutions in KYC that were just released in June 2017.

All of these upgrades, however, will move away from standard rules-based solutions, which just don’t provide the necessary level of speed and efficiency anymore for financial services — especially not in regtech.

“it’s really about, how do you enable focus on the 25% of time [firms] are spending on decision-making, and reduce the time spent on [data] aggregation?” said Andrews. “The idea of cognitive is really to build a collected base of knowledge.”