Brown Brothers Harriman To 'Transform' Its Fund Accounting Service by 2021

By 2021, Brown Brothers Harriman’s (BBH) hopes to have radically transformed its fund accounting process end-to-end. At the core of this movement is AI, including machine learning and natural language generation (NLG).

Starting last year, BBH began incorporating supervised machine learning in order to strike net asset values (NAVs) for clients. The algorithms replaced hundreds of manual-work hours by analysts, who would sift through thousands of pricing anomalies, or exceptions, finding that as low as 1% were true exceptions. Through its machine learning-powered Anomaly NAV Tracking System (ANTS), the bank has eliminated 90% of false positives.

The next big venture began in Q4 of 2019 and will be implemented by the end of Q2 this year. The system, dubbed Guardrail, expands upon ANTS and entails looking at the overall portfolio of a fund against a benchmark, such as the S&P 500 or another index, and making sure there is not dramatic price movements from a client’s fund versus the desired benchmark. Clients will be able to receive custom reports based on these comparisons.

One of the benefits of Guardrail, says Michael Cook, vice president of investor services, is the time saved by BBH’s  analysts who, if there is such a deviation, currently have to dig through all the components of a NAV to determine why it deviated, and write up an analysis.

“2021 is where all these AI projects we’re doing in fund accounting come together,” Cook says. “ANTS right now is [about] security pricing, but as we build out for the AI to review interest, and to do standardization, and to go through all these journal entries to be able to summarize and provide an analysis at the end of the day, we think that’s going to be transformational.”

The firm is also leaning on NLG technology in order to lessen analysts’ reliance on Excel spreadsheets to make decisions. After taking in all of the information within their spreadsheets—the bank has already developed its natural-language processing capabilities—the bank is building a platform that will be able to automatically generate language that says definitively, for example, “Yes, your Class A share diverged [due to an expense posting, for example],” says Kevin Welch, managing director of investor services for BBH.

Finally, the institution is also looking to develop an algo-validation tool that can help clients to more easily verify their own proofs-of-concept.

“Something like 70% of AI algorithms that are being built never go into production,” Welch says. “If BBH has built something and you’ve got a hypothesis, we have the opportunity to let [the client] leverage that to make sure it’s a good investment for them before they move forward, or they can make the decision that maybe it’s even something they want to use daily in our algorithm as well.”

The tool has been in development since July. While the service has not had client uptake yet, the bank has had preliminary talks with some interested parties.