Why Guardian Capital has put AI inside its new management strategy
“We’re a little bit late to the party with this ETF launch, probably about two to three years, in my opinion… We are not momentum managers, like every other fundamental manager is buying memory compute six months ago. And believe it or not, they’re blowing up in front of us right now. I know it. On the long only side. So not only are the growth managers blowing up, but even dividend managers who bought all these memory compute stocks are giving up a lot of alpha for us,” Iyer says. “The evolution here is we use AI to manage money. But obviously we have a deep insight into the supply chain and the evolution cycle of AI companies. And so for us, the question was, how do we create an opportunity set for ourselves and our clients? Because we are well known for dividend growth and earnings growth. And the new growth is now what we call the AI economy.”
How AI agents inform AI investing
The new Guardian ETF outlines two core groups of companies related to AI in its strategy documents: AI enablers and AI adopters. On the AI enabler side are the vast array of AI compute, memory, semiconductor, manufacturing, and equipment companies. That also includes some of the AI provider companies, many of which fell into the so-called ‘hyperscaler’ category. Among AI adopters are the companies now emphasizing AI in their corporate filings, communications, media opportunities, and operations. Given the incredibly far-reaching nature of this theme, one of the core jobs of the AI tools backing up the investment team for this ETF is in determining how central AI is becoming to a company.
Making that determination, Iyer says, is where the probabilistic mode of thinking that AI uses can shine. It’s also a prime use for AI as a collator of information. Determining the centrality of AI in a business requires discussions with management, reading regulatory filings, reading news, and comparing all that information to the businesses’ competitors. While AI may not be able to directly speak with managers, it can collect and synthesize information from those meetings, filings, and press reports. From there, Iyer says, his team built a framework of AI agents to categorize that information and use it as a base layer of evidence to protect against hallucinations. Then another layer of agents reads that information and tells the management team how central AI is to that business. The human managers then test the AI’s thesis, forcing it to check itself and, if necessary, tracing the path of its data inputs to the premise of its argument.
Once the AI has built a database of AI-centric companies, the management team uses yet more AI agents to identify the names most likely to grow their earnings. Rather than the deterministic Gaussian mathematics that certain quant models use, Iyer says that the AI agents’ Bayesian model allows for more variance because it focuses on probability. Because there is no certainty in investing, the thought process that informs an approach should be built on probability rather than certainty.
AI agents as the future of investment management
While this layered multi-agent process is being used to capture AI as an investment theme that spills over traditional boundaries of sector and asset class, Iyer believes the approach can be applied to a host of different investable areas. He believes that this approach, or versions of it, will become more commonplace.
