June 13, 2019

AIMCo Adds an AI AlphaLayer

The Alberta Investment Management Corporation (AIMCo) announced it has joined forces with AltaML to build artificial intelligence (AI) and machine learning (ML) applications within the investment management space.

AIMCo Adds an AI AlphaLayer

This article was originally posted at by Leo Kolivakis.

The joint venture, known as AlphaLayer, is designed to leverage AIMCo’s expertise in investment management and AltaML’s AI technologies and development capabilities:

As one of Canada’s largest and most diversified institutional investment fund managers, AIMCo’s move to accelerate the adoption of innovative AI/ML solutions is aimed at creating tactical intelligence to inform better decision-making and streamline operational efficiency.

“Innovation is an integral part of our strategy as we look for new ways to deliver exceptional value to our clients. We are excited about this collaboration and the possibilities – whether that be enhancing returns, improving risk management or boosting productivity,” said Kevin Uebelein, AIMCo’s Chief Executive Officer.

By providing a cutting-edge strategy and a trained data science team, AltaML offers a way for AIMCo to build AI/ML solutions at a rapid pace in a highly competitive market.

“We believe there is tremendous opportunity in combining the scale of AIMCo with the nimbleness of a startup. We've already identified numerous use cases with the potential to generate meaningful value through lowered costs or enhanced investment returns. At the same time, the partnership creates opportunities to attract and develop talent in Alberta’s growing tech sector,” said AltaML CEO Cory Janssen.

In her article, Yaelle Gang of the Canadian Investment Review notes the following:

This joint venture will leverage the AIMCo’s expertise in investment management and AltaML’s technological and development capabilities. The partners are working on use cases, or mini projects, in both the front office and on the operations side, looking for efficiencies and potentially alpha, says Michael Baker, senior vice-president of strategic implementation at the AIMCo.
Although it’s still early days, the AIMCo believes — in addition to using the developments itself — it will also be able to commercialize some of the intellectual property, he says. “That’s ultimately a goal, but we’re very early days in this.”

The AIMCo has already successfully completed one pilot project with AltaML, which makes the process for handling capital calls and distributions more efficient, says Baker, noting the process has gone from about 15 minutes to less than one second.

Now, as part of the joint venture, the partners are working on various other use cases. “The key is to be able to go in and, when it’s working, really double down, and when it’s not, be able to kill it fast and move on,” says Cory Janssen, co-founder of AltaML.

This isn’t a formal investment by the AIMCo, but rather a partnership. In Baker’s view, the approach is unique. “A lot of companies are trying to do this on their own. They’ll go out and hire a bunch of data scientists and they’ll try to do it internally. And that may work. But we just think the approach that we took — by partnering with a company that does this for a living — we think we’ll get there faster.”

The partnership component is key, agrees Janssen. “I think the big takeaway there, from the perspective of this partnership, is that you can have all the data scientists in the world, but unless you have, No. 1, the data, and No. 2, the domain expertise, there’s nothing you can actually get done.”This afternoon, I had a chance to speak to Michael Baker, AIMCO's Senior Vice President of Strategic Implementation. I want to thank him and Dénes Nemeth, AIMCo's Director of Corporate Communications for setting this call up on short notice.
Prior to joining AIMCO in 2010, Michael was at ATB Financial Services where he ran retail financial services and before that, he spent 19 years at TD in various roles. He has extensive experience and told me he will be retiring from AIMCo next year, so I was lucky to chat with him, my first conversation with a senior operational person at a Canadian pension fund (operations never get the respect they deserve, my bad).

Michael told me he embarked on this project two years ago and that Edmonton is home to the Alberta Machine Intelligence Institute (Amii) which houses some of the world’s top talent in machine intelligence, like professors Rich Sutton and Michael Bowling.

Interestingly, Michael met AltaML CEO Cory Janssen through Dénes Nemeth who had relationships with the innovation community in Edmonton and introduced them.

AltaML had no domain expertise in pension management but they are AI experts who find solutions using data. AIMCo provided the data and AltaML provided the solutions to their needs.

That pilot project which made the process for handling capital calls and distributions more efficient, reducing the time from about 15 minutes to less than one second, worked out very well.

Michael told me the process used to be cumbersome requiring a finance analyst in accounting to check a dedicated email inbox where capital calls and distributions were sent, sort out the capital calls and distributions, then fill in a top sheet.

He said they freed up resources from doing tasks that were of less value to do higher value jobs for internal clients (front office). He said they still check to make sure everything is running smoothly,  but the process is a lot less cumbersome and the machines are more accurate than humans for such tasks.

I asked him if machines are taking over AIMCo and he said he doesn't see it that way, "the machines are freeing up time so people can work on higher value jobs." But he added as assets under management grow, it's true that AIMCo won't need as many people in certain jobs which are being automated and it will be less costly and more efficient to manage these assets.

Discussing this joint venture with AltaML, he told me hiring senior data analysts who can lead a team isn't easy and it was wiser to farm out this project to a dedicated local AI-ML company that has the requisite expertise to deliver the results they were looking for.

I told him one of my buddies is a data scientist and he said these groups have to have the full backing of the CEO or else they typically run into problems with the IT department who perceive them as a threat.

Michael told me AIMCo's CEO Kevin Uebelein was 100% committed to this project and he was a senior person spearheading it.

He said they aren't there yet in terms of commercializing this process but can easily see how they might commercialize an IP on operational efficiencies.

We are literally living through an AI revolution in asset management and all other sectors of the economy.

Last week, I read an article on how asset managers are becoming data-driven in their analysis of execution quality, and they are reallocating order flows among their broker-dealer relationships at a faster pace:

Some buy-side firms are developing metrics for their internal trading desks before they evaluate their brokers.

“What we’re trying to do is align our trading strategies with our PM intentions,” said Curt Hurl, Director, Global Trading, Ontario Teachers’ Pension Plan (OTPP), who explained how the fund is profiling portfolio managers to inform its transaction cost analysis (TCA). “We try to dig into what is the intention of the PM when they send over a trade,” he explained.

“If they’re sending over a basket with regards to a quant program, is there any short-term alpha built in, or is [the trade] alpha agnostic either one or two days after the trade or during the life of the trade?” asked Hurl.

Historically TCA was implemented to ensure compliance with best execution obligations. “But, as the quality of TCA products has improved, it’s being used to inform trading desks to a much greater extent both for internal buy-side trading desks and in dealer relationships,” noted Monahan.

In the case of OTPP, TCA is used holistically to assess absolute trading costs across all asset classes at the parent level. “We use various benchmarks to measure these costs,” said Hurl. He said that OTPP is working on TCA platforms internally for futures and it relies on a third-party vendor for equities. Specifically, the fund utilizes arrival price as its main benchmark to estimate slippage. “TCA is used to measure slippage at the parent level, but we try to dive deeper into child order placements as well,” said Hurl.

“Part of what we look at is the liquidity of a stock, how do we source liquidity, is the stock illiquid, did we use blocks, did we use algos, which then goes into measuring that market impact,” explained Hurl. In addition, it looks at reversion of the stock price using time horizons — whether that occurs in seconds or minutes. It also looks at specific metrics around child placement to build out an overall metric.

At this point in its process, the pension fund is broker-agnostic because it’s trying to figure out how to trade a certain stock or basket of stocks based on the PM profile, said Hurl. “Then we could put brokers in that horse race based on similar types of strategies that we are looking for,” he said.While OTPP internalized this project, AIMCo is the first to my knowledge to partner up with a leading local AI firm to improve its operational efficiency.

Michael Baker told me he thinks the Caisse is in discussions with some local firms and this would make sense since Montreal is a global hub for AI companies.

Lastly, when I asked him why AIMCo is doing this project, he simply stated: "We can't afford to be left behind, we need to keep up with these technological advances."

He's right and I would add it's within the fiduciary duty of AIMCo and all of Canada's large pensions to manage assets in the best interests of their contributors and beneficiaries and that entails lowering costs while improving operational efficiency.

As the operations of these pensions become ever more sophisticated, they need to make sure they have the competitive edge in all facets of their operations and AI will help them achieve this edge.
Chad Langager
Chad Langager
Follow us on: