Industry faces 3 challenges to AI integration

AI has the potential to completely revolutionize the investment industry, says Greg Gipson, managing director and head of exchange-traded funds at CIBC Global Asset Management. But integrating it into ETF portfolios involves three key challenges.

“AI has the power to transform ETF strategy,” he said in a June 20 interview, helping to define what goes into an ETF, how portfolios are put together, and how risk is managed and monitored.

Listen to the full conversation on the Advisor To Go podcast, powered by CIBC Asset Management.

According to Gibson, AI will power portfolio construction in ways that were previously unimaginable.

“AI is able to process large amounts of data, unstructured data and alternative data sources, which just really enables a richer and more contextual approach to selecting the components that go into an ETF,” he said.

And while AI can be cost efficient and could help businesses scale more efficiently, Gipson said, there are still some hurdles to clear.

First, there needs to be a structured database that AI techniques can access, he said.

“As important as the machine learning or AI-based approach is, I would argue that even more important is data acquisition, data cleansing and data storage,” he said.

Second, it’s important to understand how the AI model analyzes data, Gipson said. “Without understanding what the model does, interpreting the output can both be challenging, and also lead to incorrect assumptions about what is being recommended.”

To combat this challenge, there is a burgeoning field called XAI — or explainable AI — that would explain the rationale of AI recommendations.

A third challenge involves the implementation costs to acquire data, he said.

“The actual software or processes to run these types of analysis is increasingly commoditized,” he said. “But the cost upfront to be ready to use those types of techniques should not be underestimated by any business or any user.”

Opportunities in ETFs

There are several opportunities for investors who want to benefit from the use of AI in ETFs. He described the easiest path as simply to invest in an ETF that is focused on companies that utilize AI, Gipson said.

“Think of these as thematic ETFs, where an AI or machine learning-based process is able to determine those securities that have a particular correlation or particular exposure to something like AI or data centres,” he said. “That understanding allows the ETF manufacturer, ETF manager, to create a vehicle — an ETF — that then is offered to investors to gain exposure to an area of the market that they may otherwise not be aware of.

Another opportunity is around data processing, he said.

“There are massive, massive increases in the amounts of data. It’s often fragmented, it’s unstructured, it’s alternative, it’s sitting in spreadsheets or PDFs. And really what AI allows is that automation of data consumption, and then also a clean and efficient and structured way of analyzing fragmented data.”

This is particularly important in areas where information is sparse, like when considering emerging market conditions or commodities.

“Often this data sits in an environment that’s not necessarily conducive to a systematic review or incorporation of the data into a process,” he said.

Gipson said despite the uncertainties, the AI future is bright.

“In my opinion, the outlook for integrating artificial intelligence into ETF portfolio construction is truly exciting,” he said. “The next wave of really building and developing, curating unique solutions for investors lies in the ability to leverage artificial intelligence, leverage the power of machine learning to create a more customized solution that meets individual investor needs.”

This article is part of the Advisor To Go program, sponsored by CIBC Asset Management. The article was written without input from the sponsor.