Putting AI and machine learning to work in the trade cycle
Almost anyone would love a solution to automatically sift through the emails that pile up in their inboxes every day. For AIMCo’s Derivatives Operations team, those emails are a crucial part of the work they do. While they can’t function without them, they’re keen on establishing a better way of managing them.
The group is getting support in that goal from AlphaLayer, the joint venture involving AltaML, an Alberta-based artificial intelligence (AI) and machine learning product developer, and AIMCo. When AlphaLayer began searching for project ideas at AIMCo, the Derivatives Operations team was quick to step forward with a use-case scenario.
“It was a no-brainer,” said Yakesh Chavda, a senior analyst on the team.
There are three workflows the team is responsible for — affirming trades, confirming trades and settling trades. Each workflow originates from an email with a counterparty, amounting to roughly 450 emails a month.
AlphaLayer has made significant progress on a program that identifies which of the three categories an email falls into. Although the task of classifying the emails is simple for a human, it’s repetitive and if it could be automated, the next step in the workflow could be triggered almost instantly.
“For a trade affirmation or a trade confirmation, the program will take a look at the economics presented and try to cross-reference it to the trade that we’ve booked in our system,” said Sam Ling, Derivatives Operations Manager. “For payments, the same sort of thing happens but it would cross-reference to upcoming cashflows that our system recorded.”
“From an AI perspective, this is a great use-case for automation as the underlying tasks are clear, repetitive and leverage data to be completed,” said Chad Langager, General Manager of AlphaLayer. “The past history of emails and resulting trades allowed the team to hit the ground running on training models.”
The technology’s potential beyond classification and cross-referencing is even more intriguing. The team can envision how the technology might identify errors and save time in the effort that goes into fixing them.
For example, in the payment confirmation workflow, you could have a scenario where AIMCo is expecting to pay $1,000 but a counterparty was expecting to be paid $10,000.
“Not only would the system do the first steps — classification and cross referencing — but it would start to look at some of the possible reasons for the discrepancy,” explains Ling. “It would look at all the different inputs as to how the amount was calculated, and use data to assess who likely has the right value and then indicate what area, department or data point has the potential error.”
There’s still plenty of work to be done on fine-tuning and evolving the program, but the team can’t help but look ahead to a day when some of their time is freed up and errors are reduced. When asked what the group would focus on instead, Ling didn’t hesitate.
“We would concentrate on more value-add activities,” he said, already speculating about how his team might better support AIMCo’s traders or conduct research to support new strategies.
As for AlphaLayer, the work with the Derivatives Operations team is getting closer to the sweet spot. Once proven effective and accurate, the program could be commercialized and made available to other investment managers. Profit from that becomes additional value-add for AIMCo’s clients.
“We are not only excited to see this technology fully adopted within the Derivative Operations team but will be looking to expand it to handle more processes within trade operation workflows, extend the tool outside of derivatives and start putting it in front of other investment managers as a commercial solution,” said Langager. “We feel that this initial work is the building blocks of something much bigger.”