Brown Brothers Harriman Automates NAV Reviews with Machine Learning

Brown Brothers Harriman has brought greater efficiency to its net asset value (NAV) review process with supervised machine learning, said Kevin Welch, the firm’s managing director for investor services, who was speaking at this year’s Waters USA event.

Securities pricing is reconciled each day at market close to make sure that the NAV figure is accurate, and the prices are reviewed to discover any significant variation day to day. Welch said the process, when performed with traditional methods, resulted in a high proportion of exceptions, most of which were not true anomalies, but that nonetheless had to be reviewed by analysts.

We have gone from 10,000 exceptions every single day that analysts needed to review, to under 500. We have eliminated 90% of the false-positives.

Kevin Welch

“And historically, the way most firms have done that is there was a block threshold: if a security moved by X percent, it generated an exception, analysts reviewed that exception and then either resolved it or identified that it wasn’t really an exception. And we found that we probably had 10,000 of these false-positives every single day that we had analysts going through, and it was really only a small amount that were true exceptions,” Welch said.   

To address the issue, BBH created a tool that uses supervised machine learning and predictive analysis to show how a security has moved against 800,000 others historically. “It will only generate an exception when the price is truly moving. So we have gone from 10,000 exceptions every single day that analysts needed to review, to under 500. We have eliminated 90% of the false-positives. This has been a key tool for us,” Welch said. 

Welch is responsible for workforce transformation and shared services disciplines within the BBH Investor Services division. As such, he manages a team that implements business and tech solutions, and is responsible for developing relationships with outsourcing partners.

• READ MORE: Earlier this year, Welch sat down with WatersTechnology to discuss an array of AI projects at the bank. Click here to read more. 

Welch said BBH has primarily looked at what uses emerging technologies could have in the back and middle office. “We are in the asset servicing business of asset managers, so I think when we are talking about the use cases for AI, we are hearing a lot about trading, a lot about KYC/AML, compliance. We were looking to apply AI more in the back-office and middle-office space, which I think has been under-served by AI,” Welch said.

In this process, BBH has steered away from robotic process automation (RPA). “One of the things that we realized was that RPA is pretty brittle technology. You have to change it if you have a variable process: if you change the input or there is a change in the output, you are constantly changing the code and changing the RPA. So we want to look at where to apply AI and machine learning, not just to automate processes, but also to improve the decision-making of our specialists within our business. And I think that is a much more sustainable change than RPA, which to us was a sort of short-term fixed-automated process,” Welch said.  

Growing the Team

Around two years ago, BBH started looking at use cases where AI and machine learning might help automate tasks across the firm. Initially, the bank did not have the in-house expertize to build out these capabilities, so consultants were brought in. However, BBH now has an internal innovations team of about 35 technologists and businesspeople, Welch said.

He said the main impediment in rolling out these solutions across the business is cultural. In the case of the NAV anomaly tracker, for example, analysts’ jobs changed after the tool was implemented, and required a shift in their mindset. “Their job previously was to understand finding clear exceptions on a daily basis to understand price changes on a critical security. Now they need to do that and still have all the subject matter expertize around that. But they also need to understand that they will train the machine,” Welch said. “Fifty percent of their job now on a daily basis is using the application we built to provide feedback to the machines. That was a change in mindset—it’s beyond automating the existing process but also thinking differently about the process.”

Being able to replicate these solutions across the business at scale has helped to drive their acceptance in the business, Welch said. He said he looks to solve business challenges that are common across the firm, not those experienced by one office or silo. His team will start off small with an implementation to show the organization what can be achieved, before rolling it out on a wider scale across the business.

They were trying to apply RPA and they were struggling to get the RPA to work because it’s too brittle, there are too many changes, and they were constantly updating the robot.

Kevin Welch

“I’ll give you one example … We found a lot of parts of the organization had data coming into [these units]. They were trying to apply RPA and they were struggling to get the RPA to work because it’s too brittle, there are too many changes, and they were constantly updating the robot. So we used NLP to read about 10 or 15 different types of information that are coming in from counterparties, read that information and keep learning.”

Welch said this technology is the same that teaches autonomous cars to read and recognize road signs. His team realized that the issues they were addressing with this technology were not exclusive to parts of the business, but were universal issues. 

“So we implemented it in one area and then we went on a series of road shows to roll that out to the rest of the organization. And we have scaled it across, probably, 50% of the org right now. So our goal is to identify those cognitive problems, get something live that we can show folks, and then try to replicate it from there.”