UBS and HSBC robots push into bond sales frontier

Register now

(Bloomberg) -- UBS Group AG and HSBC Holdings Plc are bringing robots into bond sales, a corner of banking still considered by many to be off limits to the onward march of automation.

UBS has created a machine-learning algorithm that shows its salespeople the most likely counterparty to buy or sell a bond. HSBC is developing a bot that will send asset managers suggestions for bond transactions tailored to their existing trading patterns.

While some bankers view bots in sales as an ominous development for one of the last bastions of old-style finance that runs off relationships, phone calls and working lunches, their bosses say the opposite is the case. Both banks say they want to make the process more efficient, relieving staff of more mundane tasks that get in the way of making money.

Elsewhere in the industry, machines are already being widely adopted, adding to job insecurity for some at a time when banks are cutting staff to counter a weakening economy.

“If salespeople are working much faster and more efficiently, you may need fewer of them,” said Tim Skeet, a career banker with almost 40 years experience in debt capital markets. “There is a fine balance between human and machine with sales because persuasion, judgment, knowledge and human contact are important, especially in less liquid markets.”

UBS’s new system makes finding a match for a bond trade easier and quicker and has reduced the average number of calls a salesperson needs to make from five to three, according to Chris Purves, head of the bank’s Strategic Development Lab.

The Swiss bank started rolling out the tool, known as Client Scout, in May after testing it earlier in the year. It sends an alert to the bank’s salespeople informing them UBS has a certain position and suggesting the most-likely candidates for the trade, ranked by percentage probability of a match, according to Purves.

“This tool is as good as the best salesperson at knowing who to contact to get a trade done,” he said. “It brings everyone up to that level.”

HSBC’s assistant, which doesn’t yet have a name, will build on an existing bot that HSBC’s salespeople started using in January to ensure they are up-to-date on the bank’s biggest events and best-read research. The tool monitors data on what research clients read and what events they attend. It has a 92% success rate when answering questions from sales, according to Ash Booth, head of artificial intelligence in HSBC’s corporate and institutional digital team.

The new system can also compile research and trading ideas for clients, giving staff more time to work on transactions.

“This has nothing to do with replacing salespeople but rather helping them and increasing their efficiency,” the bank said.


HSBC plans to test the first version of its client system with a group of asset managers toward the end of this year, according to Sotiris Manderis, managing director in the bank’s corporate and institutional digital team.

Whereas HSBC is targeting all asset classes, UBS has focused on credit because the market is particularly slow to trade. Individual bond positions sometimes take days or weeks to shift and Purves likens the market to a “dating agency” where sales try to match up buyers and sellers.

Electronic systems are making leaps in corporate bond markets. UBS trades more than 13,000 bonds electronically today, an increase of more than 20% from a year earlier, according to Purves.

Still, most of the market trades by voice and even after a salesperson uses Client Scout, the trade is executed manually between two people rather than automatically, he said.

“Machines never get bored or tired, they just keep going,” said Purves. “But the last mile is human and it’s going to be a while before that changes.”

Bloomberg News