Automated technologies like artificial intelligence (AI) and robotic process automation (RPA) can help banks and financial institutions in their anti-money laundering (AML) activities due to the massive amount of data to sift through.

“The problem we’re here to try and solve is financial crime,” said Patrick Dutton, senior vice president, head of financial crime risk assessment and head of compliance analytics for $2.98 trillion HSBC. “Artificial intelligence has been … very useful in dealing with that, because we have a lot of data. We have customer transactional data, we have investigative outputs or all those [transaction monitoring] alerts and investigations.

“We can apply a number of these techniques,” Dutton said Monday at the Association of Certified Anti-Money Laundering Specialists‘ (ACAMS) AML & Anti-Financial Crime Conference in Las Vegas.

Possible use cases Dutton suggested for machine learning (ML) tools include transaction alert adjudication to help control false positives and risk escalation, which he noted could be made more efficient. RPA can be applied to manual investigative processes to “help explain machine decisioning to investigators” by “writing down why it came to certain conclusions,” he added.

Explainability is key to successful deployment of AML automation solutions, Dutton said.

“You could use AI for today’s problems; you can use it to deploy tomorrow’s solutions,” he said, noting that today’s AML challenges are rules-based transaction monitoring, while future solutions might address front-end detection as a way to catch financial crimes sooner.

The challenge for financial institutions, Dutton said, is that they have “all the data in the world,” both customer and transactional, but lack context for it. “For deploying tomorrow, this is really about how much data and context you can bring to bear on your customer,” he said.

Common misconceptions

Financial services providers may believe they need considerable technical expertise to work with AI in AML activities, said Khalid Hossain, audit director for global financial crime prevention audit at $1.09 trillion UBS.

“There’s a common misconception that AI is pretty much out of reach for people who don’t have, let’s say, advanced mathematics or data management background,” Hossain said. “However, based on my experiences, and currently working on some of the projects that I’ve been involved [in] … [it] doesn’t require a deep level of technical capabilities. If you have them, great; if you don’t, that shouldn’t be discouraging people to … use and apply those [technologies].”

Any financial institution can use “off-the-shelf” AI technologies and apply them in AML projects and initiatives, Hossain said. He pointed to natural language processing, which can be a great tool for auditors to analyze data points within large volumes of unstructured data.

“This can become a very powerful tool if you’re working in, let’s say, AML alert handling or investigations audit,” Hossain said.

Melissa Leeds, vice president for global AML risk and scorecard metrics at $1.69 trillion Citibank, underscored Hossain’s point about technical expertise using her own experience in AML.

“I was offered this position and they said, ‘You’re doing forward-looking metrics ― okay, go,'” she recalled, noting that her knowledge of AI and ML at the time was “vague and general.”

Leeds said she taught herself the technology and how to code before finding existing tools within the institution that could prove useful. “In my free time, I’m coding; in my free time, I’m playing with our own data and building the prototype,” she said. Leeds also noted that sometimes, data scientists by trade may not be what FIs want for automating AML.

“Hiring a data scientist is not bad, but the problem with that alone is that the data scientist doesn’t have the knowledge of money laundering in general, or fraud or sanctions,” she said. “Data scientist language and money laundering language are very separated, and it’s hard to come together sometimes.”