Earlier this week I was having a debate where a bank executive was asking me about what the hottest things in FinTech are right now.
My answer was machine learning (AI) and blockchain. There are others such as contextual commerce, voice recognition and digital assistants, but machine learning and blockchain are top of mind.
His response was that machine learning was higher priority than blockchain. It was almost like if you had to invest in one thing, then the one thing he would choose is machine learning. It was like an either/or discussion, where I tried to turn it around into a discussion of both. My belief is both areas should be a priority for a bank. Why would you ignore blockchain?
His answer intrigued me. His answer was that blockchain is not essential right now as it’s a nascent technology with no proven usage. The bank can wait for a while before investing in it, and can catch up fast when needed. However, machine learning could give the bank immediate benefits in improving workflows and processes. Machine learning could save costs and overheads and was right here, right now.
I agree with this view, but disagree with the idea of waiting on blockchain. Why? Because blockchain is going to drive an essential rethinking of our industry structure. If you wait and watch, you miss the chance to guide those industry structures.2 Would you rather lead change or follow the leaders who changed everything? However, it seemed obvious that he wasn’t going to change position and, bearing in mind that I’ve said it will be a decade before blockchain is mainstream, maybe he’s right for his bank.
It made me realise that I haven’t discussed machine learning enough on the blog as well. I’ve given a few deep dives:
But I haven’t focused upon the real financial areas of usage, which from my limited knowledge is primarily in improving service to high net worth (HNW) clients. I say that as the only early discussions I’ve seen publicly about machine learning in banking has been the use of IBM’s Watson in DBS and UBS. Both banks have talked about how Watson can analyse the client’s portfolio of investments and risk appetite, so that when a HNW client opens their app, they get a personal recommendation in real-time.
DBS in particular is notable, as they’ve won quite a lot of awards for their innovations including world’s best digital bank. And no, this is not an advert for DBS, but it is notable that their CEO, Piyush Gupta, started his professional life gaining responsibility for IT, Strategy, Fiknance and Human Resources as Chief of Staff in Citibank India . He understands the power of tech.
Anyways, I was trying to think of other areas we could talk about machine learning in banking, and there are actually loads. Here’s just a few examples:
From an IBM report:
A team of researchers in the Machine Learning Technologies group at IBM Research – Haifa are taking fraud prevention and detection to a new level with the IBM Detecting Fraud in Financial Transactions solution.
Rather than singling out specific types of transactions, the solution analyzes historical transaction data to build a model that can detect fraudulent patterns. This model is then used to process and analyze a large amount of financial transactions as they happen in real time, also known as stream computing.
Each transaction is given a fraud score, which represents the probability of a transaction being fraudulent. The model is first customized to the client’s data and then updated periodically to cover new fraud patterns. The underlying analytics rely on statistical analysis and machine learning methods, some of the same techniques as used by IBM’s Watson.
From an MIT Technology Review:
USAA, just one early adopter, has been testing ways to use AI to fine-tune its detection of identity theft. Its system looks for patterns that don’t match a customer’s typical behavior and identifies those anomalies even on the first instance, Welborn says. Traditional systems wouldn’t catch a new pattern of crime until the second time it happened. “Our learning systems are really good at understanding things that look like fraud,” he says.
Another project being tested at USAA tries to improve customer service. It involves an AI technology built by Saffron, a division of Intel, using an approach designed to mimic the randomness of the connections made by the human brain. By combining 7,000 different factors, the technology can match broad patterns of customer behavior to that of specific members, and 88 percent of the time it can correctly predict things like how certain people might next contact USAA (Web? phone? e-mail?) and what products they will be looking for when they do. Without the AI, USAA’s systems were guessing right 50 percent of the time. That test is now being expanded.
From a McKinsey white paper:
In Europe, more than a dozen banks have replaced older statistical-modeling approaches with machine-learning techniques and, in some cases, experienced 10 percent increases in sales of new products, 20 percent savings in capital expenditures, 20 percent increases in cash collections, and 20 percent declines in churn. The banks have achieved these gains by devising new recommendation engines for clients in retailing and in small and medium-sized companies. They have also built microtargeted models that more accurately forecast who will cancel service or default on their loans, and how best to intervene …
An international bank concerned about the scale of defaults in its retail business recently identified a group of customers who had suddenly switched from using credit cards during the day to using them in the middle of the night. That pattern was accompanied by a steep decrease in their savings rate. After consulting branch managers, the bank further discovered that the people behaving in this way were also coping with some recent stressful event. As a result, all customers tagged by the algorithm as members of that microsegment were automatically given a new limit on their credit cards and offered financial advice.
From a Waters Technology feature:
Lauren Crossett, director of business development at Rebellion Research, considered one of the leading AI quant funds in the US, said that even though errors are made, they have to trust the AI and not override it.
“We might look at all the other funds that sold out when something starts to go down and say, ‘Ok, well we know where the fork was in the road. Let’s see how it comes out,” Crossett said. “We pretty much depend on what the AI said. It’s not to say if we think there is something wrong with the data we’re not going to rerun the system. But we’re not going to say, ‘Hey, I have a good feeling about this, so let’s overweigh something.’ We’re not going to do that.”
Erez Katz, CEO of Lucena Research, which provides predictive analytics using machine-learning technology, adds that the benefit of these platforms is that it takes human emotion out of play, which can be beneficial when it comes to execution as opposed to research.
“Machine learning takes the human’s emotion out of the equation,” he says. “That’s very important because human psychology plays an important role in most investment decisions and it often does a disservice even to the most seasoned portfolio managers. Active investors, whether they admit it or not, often buy a stock for the wrong reasons and exit under duress due to their lack of confidence in their initial decision to enter. Having a scientific backing provides the statistical affirmation and confidence in a trade, but more importantly, it eliminates the emotional angle of the decision process.”
Obviously, the list could go on and on and on, but the last example is a pretty important one, as we have long tried to replace human traders with machines. This is referred to as active (human) versus passive (machine) trading systems. With high-frequency trading (AI) and other techniques, combined with machine learning and AI, we could eradicate the need for human traders.
In a report this week in The Financial Times, active managers whose funds must try to beat the market rather than simply track the index, are facing something approaching a crisis. A majority fail to beat the index over any significant period, and most of those that do ultimately find their outperformance to be fleeting. New competitors are claiming any insight they actually possess can be replicated by a computer. Clients are shifting en masse to index-based funds — active funds have lost $213bn in assets in the year to the end of May, Morningstar says, while passive funds took in $240bn. Profit margins, traditionally among the best in the finance world, are under threat and it seems only a matter of time before there is pressure on managers’ pay … only 15 per cent of active managers are persistent market-beaters.
That is a good reason to get rid of humans in trading. According to a recent Tabb Group report, computers will replace humans entirely in the trading room, because the skin and blood brigade are expensive and prone to error. Financial market participants currently spend more than three times more on people than they do on hardware, software and data, according to the report. This does not mean we end up no humans in trading however, as there will be new jobs created for those who can build and control the technology.
In another report by Aite Group in 2014, foreign exchange (FX) trading accounted for 20% of the markets in 2001, 66% in 2013 and rising to 76% by 2018. About 81% of spot trading — the buying and selling of currency for immediate delivery — will be electronic by 2018.
We can see this shift to electronification of markets across the gamete of everything from FX to equities to structure products to wealth management to advice to service and more. However, with all this change there may still be a place for human traders that can beat the machines. It’s a different game though.
In another Financial Times discussion of the area, they quote Josh Brown of Ritholtz Wealth Management and The Reformed Broker blog. At certain times, humans have the advantage, he says. “The new game for traders is not running away from rapacious algos.Instead, it’s going to become about exploiting their failure to reason.”
Monday August 24 was one such day. Nasdaq opened limit down, partly because traders were so spooked by a plunge in Chinese shares the night before. There was virtually no volume, and individual shares plunged with no news to account for the fall. KKR was down more than 60%, mighty GE down 20%. As volatility spiked beyond the ranges they had been programmed for, the high-frequency traders, the so-called robo traders and the stat arbs did what they were supposed to do, they unplugged and stopped trading. “Volatility shy trading programs have been programmed to de-risk when prices get wild, period,” the Reformed Broker blog notes. On August 24 “it paid to ask questions first. Unfortunately, that function isn’t in the code.”
Those who did ask questions would have seen that it was impossible for GE, say, to be down 20 per cent when some index products were down far less and could have made a killing exploiting that discrepancy. Similarly, by using better data it is also still possible for humans to make money.