UBS Explores AI for Trading Functions, Real-Time Transcriptions

UBS is exploring the use of artificial intelligence (AI) to help automate front-office functions.

The Swiss bank is in the final stages of a proof-of-concept, working with Cloud9, a provider of trader workflow and communication technologies, to develop virtual turrets and offer real-time transcriptions of conversations between traders and clients.

Following a UBS breakfast briefing on innovation on September 10, Chris Purves, head of the investment bank’s strategic development labs, said that the technology will use natural-language processing capabilities to transcribe voice from multiple languages into text.

Cloud9 specifically uses a proprietary dictionary that covers terminology specific to different asset classes. The provider uses custom-made voice-translation algorithms that allows it to identify keywords in a conversation.

“The unique thing with Cloud9 is that they have what could be referred to as a ‘dictionary per asset class,’” Hyder Jaffrey, head of strategic investment and fintech at UBS, tells WatersTechnology. “The language within an asset class can be very technical and specific, something you don’t get in general transcription services, so one of the aspects that [Cloud9] offers are these dictionaries that are effectively aligned to asset classes.”

For the front office, the virtual turret is expected to free up desk space and remove legacy practices such as using telephones. If the project proves successful in helping to automate trading functions, the bank will also look to leverage the real-time transcription capabilities for other applications, such as trader surveillance, client relationship management, and sales.

“The technology will transcribe conversations within milliseconds, which is obviously very quick,” Purves says. “So, if a system can very quickly understand what is being said on a call, then we can clearly and automatically put the pertinent information that is happening in the conversation in front of trading and sales.”

The project began development in February in UBS’s UK-based innovation lab. Purves says the project was kept separate from the bank’s core systems, which allowed it to develop much faster. 

Over the coming months, UBS will incrementally roll out the technology on selected traders’ desks.