Robots Start Working On Investment Bank Trading Floors

For now, the computers and algorithms are developing strategies which are then validated by humans.

But one suspects it won't be too long before the robots are second-guessing the humans. JL

Martin Arnold and Laura Noonan report in the Financial Times:

They are putting artificial intelligence to work among their star traders, allocating funds and analysing data to develop strategies. Artificial intelligence comes up with a trading optimisation strategy that is then validated by humans. When the strategies were back-tested, they achieved annualised returns of 10.3%, outperforming the 6.9% returns of the benchmark. (But) most AI success seems currently about doing what people do in terms of process, but more efficiently.
 Robots are moving on to the trading floors of investment banks. UBS this week showcased how two artificial intelligence systems can help traders perform better at the Swiss bank’s futuristic new City of London office.Many of the world’s biggest banks have for years been automating manual, repetitive tasks done by support staff to save money. But now they are putting the latest forms of artificial intelligence to work at the heart of operations among their star traders, allocating funds and analysing data to develop strategies.“There has been a lot of talk about automation of the back office,” Beatriz Martín Jiménez, chief operating officer of UBS’s investment bank, told the Financial Times. “But we decided to start a conversation with the front-office guys on whether there were processes we could use a robot to do and we found a number of them.”The first example is a relatively simple, automated programme for dealing with clients’ post-trade allocation requests. The system, which UBS developed with Deloitte, scans for emails sent by clients detailing how they want to divide large block trades up between funds. It then processes these and executes the transfers.It saves time by doing a task that would normally take a person about 45 minutes in only about two minutes, while freeing investment bankers up for other tasks, such as calling clients.“The robot can pick up the email and do the whole allocation,” said Ms Martín Jiménez, watching the system swiftly process client requests on a big screen in the ultra-modern headquarters that UBS recently built for its investment bank in London. “It frees people up to do more complex tasks.”The second of its new systems uses machine learning to develop new strategies for trading volatility on behalf of clients. It examines vast amounts of trading data and builds a strategy based on learning from market patterns.UBS claims it is the first “adaptive strategy” product offered by an investment bank. It has been marketing the offering to clients for a couple of months and they have been “very receptive”. While the bank is yet to convince a client to put its money into it, it expects to secure its first contract within a few months.Several other banks have introduced artificial intelligence into their client-facing operations. JPMorgan Chase has developed a similar system based on machine learning for its equities business, which helps to analyse the best way to execute a big block trade by reading market conditions.UBS turned to Tradelegs, a New York fintech specialising in trading analytics, to develop its new system.When the strategies were back-tested by looking at how they would have performed, they achieved annualised returns of 10.3 per cent, easily outperforming the 6.9 per cent returns of the benchmark S&P Put-Right Index.     “The artificial intelligence comes up with a trading optimisation strategy that is then validated by humans,” said Ms Martín Jiménez, adding that it was likely to be “several years” before computers were let loose to execute trades without them first being approved by bankers.Goldman Sachs has housed a Quantitative Investment Strategies unit in its asset management arm for more than a decade. Morgan Stanley is rolling out machine learning algorithms to support brokers at its wealth management arm. But UBS says it is one of the first to introduce the technology into an investment bank’s trading floor.“If we are honest about the maturity of the industry, most of the AI success seems currently about doing what people do in terms of process, but more efficiently,” said a senior innovation executive at a rival bank.“Even extracting sentiment, which I have seen IBM do on a paragraph or a sentence or a story, it basically follows a set of rules that can explain why something is the way it is,” said the executive. “Even the learning bit — you feed a machine a thousand pictures of a hot dog and it will soon learn what a hot dog looks like.”The explosion of superfast, computer-powered trading outfits — many at hedge funds — has already raised concerns about links to a series of sudden, sharp movements in markets. The arrival of robots among the traders is likely to accentuate these “flash crash” worries.But Ms Martín Jiménez is confident the risks are minimal. “In the future it will start to make predictions about what will happen [to market volatility] and clients could trade on these,” she said. The rival’s innovation executive said: “We have such systems — and I am sure others do — that use machine learning to adapt a strategy to a specific client need at a specific time. These strategies require significant computing power and science to tailor, but are becoming more prevalent all the time. It is an interesting story I think.”