UBS Pilots Machine Learning in the Back Office10.28.2019 By Rob Daly Editor-at-Large
Artificial intelligence is finding a new home throughout UBS far beyond just developing quantitative trading strategies.
“Its applicability depends on the desired use, end objective, and problem it is trying to address,” Beatriz Martín Jiménez, investment bank COO and UK CEO at UBS, told Markets Media. “At UBS, we extensively use AI across the entire operating cycle.”
Advances in technologies, such as machine learning, have made it possible for the bank to reimagine services and process delivery in a holistic fashion, she added. “In the past, we may have focused on solving distinct problems for an immediate fix. While looking at the broader and more permanent solutions may be harder, it will set us up for long-term success.”
UBS has been working on proofs-of-concept and pilots within operations to process unstructured data.
Over the approximate past year, the bank has worked with an unnamed vendor to develop the pilot’s code that will help the firm classify roughly one million emails that the operations department receives daily across an estimated 5,000 mailboxes.
The pilot’s goal is to identify the emails that likely will require an escalated response from the operations team, such as potential trade breaks before they happen, based on the phrasing and vocabulary used by an email’s author.
Although UBS is working with an outside vendor, it is UBS employees who identify the patterns and decide how to proceed.
The bank is happy with the pilot’s performance and expects that the pilot’s small team likely will complete the project over the next few months, she added.
One reason her team decided to implement this proof-of-concept and pilot due to the Securities Financing Transactions Regulation’s upcoming deadline, which provided the team with a hard deadline.
The pilot also provided the bank with experience with reading, classifying, and prioritizing unstructured data that it can apply to other functions within its middle- and back office, such as the legal department.
A scenario might have machine learning compare a term sheet with a client’s latest email or call and give the bank the ability to check those terms against the parameters in the bank’s risk system, which then the legal department can conduct any necessary triage, according to Martín Jiménez.
Implementing such new processes into eventual production is not an attempt to lower headcount, but to redirect employee’s efforts towards more valuable tasks and to reduce their manual and repetitive tasks, she added. “It is about achieving more and better with the same workforce, engaging our staff on where their skill and expertise is far more impactful.”
Technology has enhanced capabilities of surveilling larger and more disparate data sets.
With Eugene Kanevsky, James Redbourn, and Joanna Wong, CLSA
AI/ML on the buy-side trading desk is a long-term program with short- and mid-term deliverables.
Aiden uses the computational power of deep reinforcement learning to improve trading results and insights.
BondDroid’s AI-generated prices are integrated directly into LTX’s pre-trade analytical tools.