Maryanne Richter, executive director, electronic credit trading strategy, global credit at Morgan Stanley, said the bank is looking for a way to automatically convert chats into request for quotes in credit markets.
Richter spoke in a webinar, Unlocking Liquidity through People and Machines, hosted by consultant Greenwich Associates last week.
Brad Levy, president and chief commercial officer at Symphony, said on the webinar that the secure messaging and collaboration platform provider was spending a lot of time on unstructured data.
“We have structured chats for interest rate derivatives and equity derivatives into RFQs,” he added.
Last year Symphony launched Sparc, a pre-trade negotiation communication tool which uses bots to automate the RFQ process.
Richter warned that artificial intelligence and machine learning are not magic bullets. “A lot of the data is bad or unstructured, there are cumbersome workflows and legacy technology which cannot be easily be lifted out,” she added.
However the bank has built a natural language processing tool to analyse mortgage securitisations for the Libor transition.
“The system identifies the category of Libor fallback so our traders can quickly understand the risk without reading hundreds of pages of documentation,” Richter said.
After the financial crisis there were a series of scandals regarding banks manipulating their submissions for setting benchmarks across asset classes, which led to a lack of confidence and threatened participation in the related markets. As a result, regulators have increased their supervision of benchmarks and want to move to risk-free reference rates based on transactions, such as SOFR in the US, so they are harder to manipulate and more representative of the market.
Richter continued that another trend is bilateral connectivity between a bank and a client through APIs. “This is in its early days and needs to made scalable as it requires a lot of resource,” she said.
She added that the majority of the bank’s trading is still by voice although there has been a big increase in electronification and the use of algorithms. The bank can use algos to systematically price 12,000 investment grade and high yield bonds in the US, and 4,300 in Europe, as well as to respond to some RFQs.
“Some algos can can suggest a price to the trader,” said Richter. “We use technology to augment human skill and find more liquidity for our clients.”
Levy said the financial industry needs to look at a broader set of technologies, especially those that operate like consumer apps that are lighter and more agile, unlike the heavy infrastructure used by the industry.
“We are in the center of distributed tech, largely enabled by the cloud,” he added. “There is not one technology that will change everything but lots of things that will combine together.”
Ideal Prediction
John Crouch founded Ideal Prediction in 2015 as he felt there was a need to combine the use of technology and data with market knowledge. Ideal independently checks that trading activity in fixed income, currency and commodities abides by best market practices and regulations, and presents the results in an intelligible and user-friendly interface.
Crouch told Markets Media: “I worked in trading for 20 years and data scientists would pitch to us who had no market knowledge. Banks need independent firms with domain expertise and the right technology.”
He described Ideal Prediction as performing supervision, rather than surveillance, as the platform monitors behaviour for which there may be no rules.
“For example, a client may be treated differently in some way and if this is appropriate, the bank can document the evidence,” he added.
With staff having to work from home due to the Covid-19 pandemic, Crouch said banks are reaching out the company as they need better tools for supervision.
“Our implementation of the FX Global Code means we have the right story at the right time,” he added.
