Whether for surveillance, certain forms of regulatory reporting, anti-money-laundering (AML) or know-your-customer (KYC) compliance, artificial intelligence—including, to a growing extent, machine learning and cognitive tools—have proven beneficial.

While speaking at this year’s North American Financial Information Summit held in New York in May, compliance officers from Bank of China and Guggenheim Partners noted that there’s a lot of interest in cognitive tools, but that it’s important for firms to take their time when implementing these solutions.

Abel Picardi, chief compliance officer for Bank of China, said that the institution has started to work with IBM and is incorporating its Watson system. With that said, it is taking it slow.

“AI could be the solution, but it has to be implemented the right way. And the right way is taking the time to understand what the [business] requirements are for the financial institution,” he said. “We’re working with IBM Watson, but we’re going slowly there because we want to make sure that everything that goes into [the technology stack] we understand, and we need to explain it to our regulators. If our regulators don’t understand it, we don’t stand a chance to validate that AI system.”

• [Listen: IBM’s Marc Andrews joined the Waters Wavelength Podcast to discuss how IBM Watson is being targeted toward financial institutions. Click here to listen.]

The hope is to eventually move to a system “that captures everything,” Picardi said. Every financial institution has to navigate enormous amounts of differing regulations, and that can lead to siloed solutions spread out across the organization. Rationalizing those systems brings efficiencies and cost savings.

“Working in a silo you have to collect everything from three, four, five, six, seven systems, and then someone has to consume that data [and make sense of it],” he said. “This is where [in theory] the project with AI comes in—the AI collects all the information that’s within the bank, it’s going to put it all in perspective. Well, that’s easier said than done. The plan sounds great, the objective is ok, but the execution is another story.”

Give Me All The Data

Joseph Lodato, global head of compliance surveillance and technology at Guggenheim Partners, agreed with Picardi that AI has its benefits, but that firms need to move slowly when deploying AI to handle compliance matters.

He said that he also wants something of a one-stop-shop for all the data of the organization. “I’m compliance, so I want everything. I like to know every card swipe, every phone call, every surveillance, every trade—we want to put that all together,” he said.

That means improving data linkages and building out a case-management system that, essentially, has everything “on one single pane or window, one glass,” that has all the information needed to make a decision: should a compliance officer close the alert, escalate it or investigate it? But cognitive tools could help move that process forward and allow the institution to solve the issue long before it becomes a problem. “In compliance, we don’t want to be reactive, we want to prevent the problems because they can cost a lot of money when the regulators sink into it,” he said.

Lodato added that space is becoming crowded with solutions, as almost every major bank has some sort of incubator or venture arm, and cloud-based tools have made it easier for fintechs to reach the market. But, echoing what Picardi said earlier, vendors need to better understand the firm’s business needs. And firms, too, need to understand what their problems are before hunting for a new tool.

“The people who come in with the technology oftentimes come in with this nice shiny new thing, they say it’s great and you need it,” Lodato said. “A year later, they realize that the sales cycle is very different. These are great tools, but if I give somebody a hammer and he needs to drill a hole, it doesn’t work and I don’t care how shiny the hammer is. …There are a lot of ideas coming out—and that’s great—but at the end of the day, you have to solve a problem. These tools, all this technology, it’s available now and easy to get to, but it still has to solve a problem and as an organization, we have to know what those problems are.”