Fewer Band-Aids, more fixes: Brown Brothers Harriman unifies busy suite of AI products

There’s something about the 1920s. A century ago, the Roaring ’20s gave the world bootleggers, flappers, and the rise of jazz bands. This time around, the world has so far experienced a global plague, some really bizarre world politics, and “hybrid work”. While there have certainly been some roadblocks in navigating the last three years, it’s provided inspiration to some, such as Kevin Welch, managing director in Brown Brothers Harriman’s Investor Services unit.

“We’re thinking about the art of the possible,” says Welch, who is responsible for workforce transformation disciplines and has spearheaded the development and rollout of several artificial intelligence solutions over the last couple of years.

But 18 months ago, Welch and his team started to pivot. It was a year and a couple of months into the Covid-19 pandemic, and the office setup of BBH—and of its clients—was still in flux. Colleagues who once saw each other five days per week were still distributed across cities, states, and beyond. And, Welch realized, so was the bank’s suite of AI solutions, which included Ants, an anomaly tracking system that weeds out false exceptions during net asset value (NAV) strikes; Linc, a reconciliations tool; Guardrail, which monitors NAV movements and flags variances from historical performances; and an algo validation tool, which enables business users to verify their own proofs-of-concept more easily.

The formerly separate solutions, which had been deployed internally at the bank and externally for clients, now constitute a single, shared-architecture platform, dubbed SmartCAP, which stands for cognitive automation platform. It serves as a central place for model accessibility and features a user interface so that users can directly interact with the models, provide feedback, and perform data labeling.

“When we talk about decision-making automation, you have a process already in people’s heads,” Welch says. “We’ve really been thinking about how to take some of that process and decision-making that has resided in an individual’s head and digitize that.”

The theme underlying SmartCAP is novel, and it seeks to answer two questions: What happens to an organization when individuals leave—especially during turbulent times of high turnover—and take all their expertise with them? And is there a way to retain their assets even if the person is gone?

Welch offers an example—failed trades, a space that is being increasingly regulated. In 2014, Europe introduced the Central Securities Depositories Regulation (CSDR), which imposes penalties for failed trades. And in the US, there’s an ongoing move toward shorter settlement times.

SmartCAP takes in all trades BBH has in its system and puts them up against a handful of pre-set rules meant to identify high-risk trades that might be unmatched as of the settlement date. The process brings those trades to the forefront of the bank’s specialist queue, which processes them through the platform’s UI. The specialists label the resulting data into a cohesive dataset, which then starts to capture the specialists’ subject matter expertise in an algorithmic form.

“Over time, what that does is give us a dataset that can service us in two ways. One, our expertise in failed trades, with 10 or 15 years of knowledge, is now digitized. And as others perform that function, we can give hints and recommendations that allow them to think how to go about resolving this exception,” Welch says. “Then, ultimately, when you get to a certain level of confidence within the algorithm, you can use it to actually automate the transaction.”

The use-case may evoke a classic anxiety that was also born in the 1920s with the advent of mass production—the worry that robots are coming for your job. That’s not the case here, says Welch, who believes practices such as digitizing subject matter expertise is an aide to workers, particularly young ones who are not only onboarding onto their first company, but often their first industry.

Octavio Marenzi, CEO and founder of Opimas, a capital markets consulting firm, says many processes seen in operations and technology at banking institutions are not documented well, which makes for fragile systems that can easily break downstream when even small changes are made to them. But he isn’t sure that AI models, known to be unexplainable black boxes in nature, are the right way to get documentation right.

Marenzi says BBH’s approach reminded him of a story about a piano maker that had documented its instrument-crafting process down to the millimeter in terms of how they cut the wood and which tools to use. On one day, the shipment of wood they received was slightly wetter than usual, but the crafters followed the existing handbook exactly.

“The fourth generation craft expert would have known that if a slightly moister piece of wood comes in, you’d cut a bit this way, or you’d go diagonally rather than that way,” Marenzi says. “It’s hard to capture all those exceptions, so as a result, the pianos sounded terrible.”

Marenzi acknowledges that clearing and back-office functions are more cut-and-dried than piano making, but he maintains that the gist of the story is the same: Documentation is hard to do. And he thinks it’s harder to do with AI.

Joshua Fine, head of digital and data for BBH investor services, has overseen the nitty-gritty of the platform’s development. He says that in the planning of the platform and its use-cases, the team had a clearer idea of what they didn’t want to offer clients so they would leave what they could offer more open and conceptual.

“We don’t want to give our clients software and tell them to go have fun with it. We want to give them a really robust platform that allows them to pick up the phone and ask us questions, and we can help solve their problems that way,” Fine says.

One of the initial lessons learned, as BBH was going to market with the cohesive product, was that clients focus most on specific use-cases for which they want to use the data. This makes sense, of course, but it leads to the old mentality of building and rebuilding point solutions, spreadsheets, and ad hoc solutions.

“That’s always been our hypothesis in the market. And it’s certainly been our experience ourselves as we worked in our own environment,” Fine says.

Instead of solving individual problems, BBH wanted to create a flexible, well-managed, and far-reaching dataset that could serve an array of use-cases, offer more operating leverage and scale, and drive multiple outcomes—a physical manifestation of Welch’s “art of the possible” mantra.