While artificial intelligence-powered tools are being promoted as the next wave of innovation in banking, there is a deep human effort involved in their construction.
“You have to have hundreds and hundreds of conversations,” with experts from many professional fields, said Jeffrey McMillan, Morgan Stanley’s chief analytics and data officer, in a panel at the In|Vest Conference.
In the traditional way of machine learning, technologists might sift through millions of phone calls to train a bot. Since the bank does not record client calls and the daily customer issues financial advisers face are complex, Morgan Stanley has to find experts who can provide detailed answers to questions from the bank.
“You build up a corpus of knowledge,” McMillan said. “Navigating the complexities of family relationship is difficult … financial advisers want to know more about psychology.”
A long-term effort is developing AI to provide recommendations for psychological advice, he added.
But McMillan emphasized that Morgan Stanley is not trying to create a bot that communicates directly with clients, and said he believes that would be dangerous. His team instead is trying to deliver knowledge to their workforce in a way that can scale.
Another lesson learned in developing automation, especially in-house, is that staff have to be managed constructively when issues arise, McMillan said, noting it’s usually a problem with process as opposed to an issue with the bank’s machine learning technology.
“Nobody comes to work in the morning wanting to be irrelevant they want to make an impact,” he said. “What’s the workflow look like and how can I get a person to change it in the system and work that technical magic?”
McMillan solves this issue by showing his employees which of their behaviors are driving positive outcomes in the research.
“People want to be relevant in their work,” he said. “A lot of this is just providing transparency and creating the right measurement framework so you have the tools in the right place so people understand how what they do ... translates into success.”
Morgan Stanley’s new AI recommendation platform, dubbed next best action, has only been live in its current form for about two months. The platform allows advisers to send customizable messages to hundreds of their clients, at any hour of the day, giving individuals a more personalized and engaging relationship with their adviser.
“If you are a very large client, you were probably getting this service already, because you can do it for 10 or 15 people. But try it for 300,” McMillan said.
With this service, McMillan says the next time the markets go down 4%, advisers can send hundreds of clients emails with personal messages and recommendations on what to do.
The tool compresses the time it takes to find out clients’ asset allocation, tax situation, preferences and values into seconds, and then lets the advisers evaluate the provided alternatives and decide what information they will pass down, as well as the best way to do it. The information is never sent directly to the client.
The bank measures the value of its AI platform by looking at five variables: cash flow, brokerage business volume, new advice clients, the level of banking business, and account attrition.
“We take those factors and build models that measure what the lift is on the individual product level,” McMillan said.
Jessica Mathews contributed to this article.
Corrected July 12, 2018 at 3:00PM: An earlier version of this story mistakenly said that Morgan Stanley sifts through millions of calls each month between advisers and clients to help build its AI. Those calls are not recorded.