Citigroup is leveraging machine learning and artificial intelligence to reduce the burden of sanctions.

“It’s been hugely impactful in the sanction space,” said Scott Nathan, global head of detection and transaction monitoring for Citigroup. “This is a use case that any institution especially today, especially in this climate, should be considering: how you can apply this machine learning technique to really zero in on the productive output. We found extreme productivity in this space.”

It’s a use case other banks should be pursuing, especially now as sanctions are changing daily across multiple governments agencies, Nathan said during a discussion on “Brainstorming: What the Next Wave of Artificial Intelligence Means for Anti-Money Laundering (AML).” The panel was held Monday during the Association of Certified Anti-Money Laundering Specialists (ACAMS) Hollywood virtual conference.

“I don’t know a bank in this conference, or in the world right now, that hasn’t had their sanctions teams or financial service company completely inundated,” Nathan said. “I mean, people are sleeping on the floor.”

Machine learning works well for sanctions due to the extensive real-time data and risk created in stopping a transaction or having a credit crisis, he added.

“On the sanction side, you’re dealing with real time, millisecond response times; you need machines that can process and respond in ways that humans just can’t,” Nathan said.

On the fringe

The panelists agreed that automation so far has lived on the fringes of AML.

“If I could characterize what that automation typically looked like, I like to say there was a lot of automation automating around the edges of compliance processes,” said Grant Vickers, who leads intelligent automation firm WorkFusion’s financial crimes go-to-market strategy. “At the end of the day, it was like a light touch approach.”

Jack Kasperek, BSA executive director at Wintrust Financial, a $43 billion holding company for 15 chartered banks and seven non-bank subsidiaries, agreed.

Robotic processing automation has been key in automating aspects of AML at the, he said. Wintrust intends to build out that RPA framework to achieve further automation.

Change accelerates strategy shift

But the automation of AML at the fringe is shifting to more mainstream functions in part because of the pandemic and pressure created by the ongoing Great Resignation, the panel agreed.

Roughly $800 billion to $2 trillion annually is being invested in AI tools for AML, according to a survey of 850 ACAMS members worldwide about their employer organizations’ use of technology to detect money laundering.

Panel moderator Bernard Williams Jr., eBay’s head of AML compliance governance and compliance training, cited the survey during the Monday discussion, noting that the pandemic and other disruptions have meant “the migration of manual functions toward AI solutions have been greatly accelerated.”

WorkFusion’s Vickers agreed. “Banking executives are getting better at coming up with a holistic digital strategy for their compliance groups,” he said, adding “No longer can executives come up with a point solution and duct tape something to a legacy estate model and hope that that type of automation is going to hold up.”