How AIMCo is building a scalable artificial intelligence program
While many organizations are grappling with how to incorporate artificial intelligence into their investment or back-office processes, the Alberta Investment Management Corp. has found a way to do it that’s scalable and has potential for commercialization.
The AIMCo started down this journey because it realized that if it didn’t adopt AI and machine learning, it would fall behind, said Michael Baker, the organization’s senior vice-president of strategic implementation, during a session at the Canadian Investment Review’s Plan Sponsor Exchange conference in February.
Like many organizations, the AIMCo initially considered hiring data scientists internally, but decided otherwise as finding that type of talent is difficult. Instead, it partnered with AltaML, an Alberta-based software and machine learning company, to form a joint venture called AlphaLayer.
In determining which projects to focus on, the AIMCo worked with staff on the frontline to help formulate ideas and be part of the projects through their lifecycles, said Cory Janssen, co-founder of AltaML, who co-presented at the conference.
This first project — incorporating AI at the AIMCo — focused on streamlining the capital call process for the organization’s real estate portfolio. Initially, its operations accounting team had a very manual process for these capital calls, which took an individual accountant 15 to 20 minutes. However, the use of robotic process automation, natural language processing and machine learning reduced this process to seconds. “It’s really allowed us to have our people spend their time [on] higher value ideas” said Baker.
It’s key to think about standardization when creating automation processes, noted Janssen, since these can be adapted to use in other areas of the organization.
For example, the capital call platform was successfully modified for a use case involving the AIMCo’s derivates portfolio, which had a tedious process of settlement that relied on incoming emails. Using the initial platform, an automated process was designed to match the trades to the right counterparty, reducing 10 to 15 minutes of work to mere seconds.
Another example of a use case was the introduction of AI to the AIMCo’s externally managed hedge funds portfolio. Historically, it required two analysts to spend two days per month inserting data from monthly statements into Excel spreadsheets, which the accounting team further verified. Since the AIMCo built an automated system that reads information and populates these sheets, it’s able to save four employees two days of work per month.
“When you have it in a proper format, then you can start to look at the analytics that the portfolio manager can actually look at to make insightful decisions on what they want to do with some of these hedge funds,” said Baker.
When thinking about building solutions to solve problems, it’s also important to think about the potential for commercializing the solution, noted Baker.
And, once projects are identified, the AIMCo and AltaML focus on small bets and fast fails. “We’ll go in with a statement of work — super small,” said Janssen. “Almost all these initial proof-of-concepts are under $50,000.”
It’s important to identify early on if the data suggests the project will succeed, he added. “By having this staged approach, it’s really allowed us to be flexible and not go down some rabbit hole that ends up costing us hundreds of thousands or millions of dollars.”