Australia's ANZ Group has revealed that it's using Watson, the IBM computer that won a round of Jeopardy two years ago, in its wealth management division. It's the first bank to provide any level of detail about its plans for Watson, although several banks, including Citigroup and Royal Bank of Canada, have been quietly experimenting with it in their labs.
"We're always investigating new technology and how we might innovate in our financial space," says Joyce Phillips, CEO Global Wealth and Private Banking Group Managing Director, Marketing, Innovation and Digital at the $642 billion-asset bank, which operates in 32 countries. "We heard about Watson from some of the tests it was doing in the States and some of the banks that were considering how to deploy Watson."
Watson is a combination of natural language understanding (it can "read" and understand unstructured data found in documents such as contracts, letters and memos), a search engine that can comb through millions of data points in seconds, and artificial intelligence that "learns" how to draw conclusions from the data. It can accept questions written in normal English, mine mountains of documents and databases for the answers, and provide natural language responses.
"I started to think about how many apps could benefit from reducing complexity and Big Data into simple solutions," Phillips says. "You could see the implications for doing that in customer service, in underwriting and in other areas that are ripe for innovation."
She wondered if Watson could be used to assess a new customer's financial situation more quickly and comprehensively than a human being could. "What if I could sort you out in one session by asking you a few questions, and not have you bring a file cabinet of paper to me, while we're having a cappuccino potentially?" Phillips mused. "I started to test that concept with clients and it really resonated." ANZ has about 2,000 financial advisors and five million wealth management customers in Australia and New Zealand.
"Imagine if I was able to read every single financial document that exists in this country and never forget a word I read, and then we sat down with great advisors to understand how you would home in on the most relevant pieces of information and be able to do that consistently, within a matter of minutes," she says. "That's what Watson can do for us. It's more than just retrieving data or a search engine. We want this to be an assistive tool for the advisor."
A team of 40 people at ANZ are training Watson by feeding it documents and data, just as IBM trained the system to compete on Jeopardy. These include advisors, product experts, legal and compliance staff and customer service people. "We're making sure we're looking at data from all sources and how that might be retrieved and used. It's quite a big initiative," Phillips says.
Some of the data is loaded electronically, some is scanned paper documents. "The great thing about Watson is it's agnostic about the data source," Phillips says. The system is being given information about the bank's products, including their latest terms and conditions.
"A lot of making Watson deliver the experience we'd like to deliver is educating the system, loading in vast amounts of data," Phillips says.
In another phase of the training, financial advisors are submitting questions to Watson and guiding it toward appropriate answers, thus refining the tool. "A lot of what we have to do is put this in human language as opposed to legal language," she says. "We have a lot of refining to do." The training period will continue until at least December.
Watson is getting smarter with each session, Phillips reports. The hope is it will help advisors deliver advice with a consistent standard of quality and eliminate the weeks' worth of work it can take to create a financial plan for a client. "We've got a lot of work to do, but we're quite impressed with how Watson is responding already with what we're offering clients in the test setting," she says.
There are no plans to let customers interact with Watson directly; the computer might produce an answer that's wrong for that client and regulators might raise concerns.
Phillips sees this tool as a means to retain financial advisors. Leaving a bank that offers the use of Watson, "Your IQ goes down 20 points because you went across the street to someone else — [instead] you'd want to stay and serve your clients with this tool."
Once the training phase is over, ANZ will begin piloting Watson with a statistically valid sample of clients.
Advisors will connect with Watson through tablet and mobile technology. The bank is developing a digital portal for wealth management that will include online chat.
The eventual return on investment, Phillips suggests, will come from increased fees charged for advice, increased market share and improved customer retention.
Assuming all goes well in production, the bank plans to bring Watson to other business areas, such as operations and customer service.