Asked if a computer will ever be able to give better investment advice than a human, Oliver Bussmann does not hesitate.
"I believe it's possible," said Bussmann, who until March was the chief information officer of UBS.
Banks' wealth management departments and other investment firms are starting to adopt artificial intelligence. This is different from the robo-advisers you've probably heard about. Those have simplistic, rules-based models — you give them your age, risk tolerance, goals, and so on and they select a basket of exchange-traded funds for you. The next generation of AI in wealth management uses rules or models crafted by data scientists with Ph.D.s and master's degrees.
"They can come up with very sophisticated investing models, leveraging artificial intelligence technology, and potentially can outperform the traditional players," said Bussmann, who now has an eponymous advisory firm and is also a mentor at Level39, a fintech accelerator based in London. "With AI you can scan the available market data and understand events and triggers that change the market situation and potential performance of certain sectors, certain stocks. It's all about the ability to process a huge amount of data, define the rules and drive the right rules."
To be sure, overreliance on computer models has gotten the investing world in trouble in the past. High-frequency traders have at times caused volatility and flash crashes through their use of algorithmic trading. The quants (or quantitative analysts) who poured into Wall Street in the early 2000s were heralded for their risk and pricing models, which enabled firms to sell complex collateralized debt obligations to unsuspecting investors that turned out to be based on subprime mortgages. Computer models by themselves can only do so much.
For this reason, Bussmann says, it will always be important to keep humans in the equation.
"You have always two extremes," he said. "One is, you rely 100% on the models; the other is, you rely just on the human. I think it's necessary to find a balance." The fully staffed bank chief investment officer department — which analyzes market sectors, makes portfolio allocations and picks stocks — may migrate to an environment where there are people modeling data and people validating the models and their results.
Financial advisers' jobs will change as well. Over time, they will create investment strategies that are used to build the models. Then the advisers will communicate the advice the models generate. Machines will take care of the execution and maintenance of the strategies.
"The industry is already racing to maximize productivity," Bussmann said. "How many assets under management can you manage as a financial adviser? Even in the high-net-worth environment, you have to think about how much can you automate and drive efficiency while not losing that personal touch."
Competition will come down to "how good is your combination of engineers, scientists and market people to define and improve those models," Bussmann said.
Early Adopters
ANZ, in Melbourne, Australia, was among the first banks to explore the possibilities — it began using IBM's Watson to help its financial advisers understand their clients in 2013. The hedge fund Bridgewater Associates hired the former chief engineer behind IBM's Watson supercomputer in 2013 to create a unit that would use AI to predict market trends. BlackRock built its own artificial intelligence engine to assist its investment managers and clients. Goldman Sachs uses the AI-based financial research platform Kensho. UBS, Deutsche Bank and others are using an AI engine called Sqreem.
BlackRock's AI engine, called Aladdin, is used internally to help inform investment decisions; it's also offered to customers. Nearly 30,000 people use the system.
"There's information out there that is highly correlated to stock return, bond return and market return," said Jody Kochansky, head of the Aladdin product group at BlackRock. "We believe that the firms that can get organized around their data, understand and are able to research what all those data are telling them and predicting can ultimately invest in a way to create better returns for clients."
Aladdin, which is based on open source technology and
Social media feeds give Aladdin insight into news events. If people tweet photos of a fire that broke out near a gas pipeline, that could cause a scare 45 minutes later in the energy market, for instance.
"That ultimately helps the human make an investment decision," Kochansky said.
The software can gather satellite images, for instance, to see how full a retailer's parking lot is. That's likely to correlate to the company's revenue, which should correlate with the stock price.
BlackRock thinks in terms of machine learning informing human investment decisions, not yet in terms of machines making decisions on their own.
"As a fiduciary, you have to always understand why you are making a particular investment decision," Kochansky said. "We have to be careful about letting loose artificial intelligence."
One time, BlackRock's operations team wondered about a set of trades that had failed. Using Aladdin, they quickly discovered that a counterparty in Brazil was truncating prices to four decimals when the market standard was six decimals. A phone call was made to the counterparty, who addressed the matter. The fail rate for those trades dropped to almost zero.
Artificial intelligence "is allowing us to see the forest for the trees, to see better the patterns," Kochansky said. "It's allowing all of us to know where to focus the human creative element. You're able to test more hypotheses more quickly and then home in on the ones that are going to help."
Reading the News — and Clients' Minds?
Goldman Sachs invests in and uses Kensho, which applies machine learning to huge datasets of news and stock movements to answer questions such as, "How do defense stocks react to terrorism incidents in Europe?"
Sqreem (an acronym for Sequential Quantum Reduction and Extraction Model) analyzes people's digital footprints and behavior to predict which products and services they're most likely to want. Wells Fargo, BlackRock, UBS and Deutsche Bank are users.
The model's deep learning algorithm, which has been under development for nine years, doesn't have to be programmed, according to Ian Chapman-Banks, CEO of Sqreem Technologies. It finds patterns on its own.
The software has behaviorally mapped 300 million people in the U.S. as well as the populations of 40 countries. In each country it deploys hundreds of thousands of web crawlers that read every news article and analyze searches done in that country. (Google makes anonymized search information publicly available so advertisers can know which keywords they want to buy.)
"Search is pure intent: 'I'm looking for an ETF,' 'How is my ETF doing?' 'I'm looking for a fund,' 'I need a credit card,' 'How do I get a loan?' " Chapman-Banks said.
In wealth management, Sqreem has mapped the likes, interests, life stage and activities of high-net-worth individuals, including which banks they use. It finds patterns: for example, people who bank with UBS in Singapore have a high affinity for ETFs.
"We can enable any bank in the wealth sector to understand what their customers want to buy, then we can help them sell and cross-sell," Chapman-Banks said.
"If you're a client adviser and you've got 40 clients, you're managing $5 million to $10 million and they've got $2 million to $6 million at other banks, it's difficult for you to understand the total picture of that client," Chapman-Banks said. "We're able to build a rich profile of a person so we can understand what their affinities are, sell them products they really want."
Advisers get weekly tips about topics and products to talk to clients about, by iPad or mail; next year a chatbot will be offered directly for clients and advisers.
Bussmann estimates it will be another three to five years before artificial intelligence makes a big impact on the financial advice industry.
"From my perspective, we're at the beginning," he said. "My observation is there's a good chance those models will drive better performance."
Editor at Large Penny Crosman welcomes feedback at
