Payments technology company North American Bancard Holdings (NAB) rolled out an automated, machine-learning based anti-money laundering solution for its transactions this month, with plans to expand the tool’s use to other transactions in the future.

The Troy, Mich.-based NAB manages $45 billion of transactions annually for businesses and banks, handling credit, debit, check conversions, and guarantee and loyalty card solutions.

The fintech partnered for the real-time solution with software as a service (SaaS) provider Hawk:AI, which predicts a reduction of NAB’s manual due diligence processing time by up to 70% with the solution. Previously, hunting for anti-money laundering (AML) anomalies in the transaction data was a very manual process, NAB Chief Experience Officer Jim Parkinson told Bank Automation News.

“Our folks would have to use basically manual data-mining techniques to find potential money laundering situations, and then they would have to do all the research themselves, and they would have to manage the case with our ticketing system and so forth,” Parkinson said. “Now, we have the system hunting for them.”

NAB selected Hawk:AI because its solution combines machine learning with a rules-based system, Parkinson said. He rejected the idea, espoused by Google in an Office of the Comptroller of the Currency filing last year, that rules-based systems were fragile and should be replaced by machine learning-based systems.

“Machine learning is an amazing tool and it’s really powerful,” Parkinson said. “But there are certain patterns and things that machine learning just has not evolved to yet, that really require a blending of machine learning and rules, which is why we, quite frankly, picked Hawk: AI.”

NAB had already worked with one vendor on AML automation, but did not get the results Parkinson wanted. Hawk:ÁI was one of the few vendors that offered both approaches, he added.

Machine learning helps reduce false alerts while flagging transactions that might otherwise be missed, said Tobias Schweiger, chief executive officer and co-founder of Hawk:AI.

“It’s really both efficiency and catching the bad behavior earlier and in a better way,” Schweiger said.

This is the first major U.S. deal for the Munich, Germany-based company, according to a press statement. Hawk:AI, which launched in March, 2018, held a seed funding round in October for an undisclosed amount, according to Crunchbase.

NAB was unable to offer specific KPIs due the system’s early April rollout being so recent. However, Parkinson said his staff is already pushing for expanding the use of the approach. Parkinson did not say what, specifically, the next step would be, but did say once NAB makes sure it has “everything right, then we’ll start to move to the next step.”

“The good news is we’ve solved the hard problem, which is looking at all of our transactions in real time,” Parkinson added. “We’re not limited by scale or scope or any of that stuff. So now that we’ve got that working, it’s a progression of functionalities.”