Here’s AI-managed ETFs’ big secret: They have an investing mind of their own

AI programs have been thrust out of their theoretical environments and into the low-cost exchange traded funds market. But they remain a bit of a mystery even for those that run them.

“Here’s the big secret with respect to AI…we have no idea,” Steven Hawkins, president and co-CEO of Horizon ETFs, whose AI fund Mind currently holds $15 million in assets, told the Financial Post.

Mind, which harnesses millions of data points to pick stocks, is merely told “these are all the things that you should think about,” Hawkins said. It’s fed millions of news articles as well as quantitative data, such as average price-to-earnings ratios and the historical volatility of that index. But the program develops its own decision-making algorithms.

“We haven’t told it, say, ‘use a momentum-driven strategy to make your asset allocations,’” Hawkins added. “We cannot predict what the AI system is going to do.”

While Mind is not tied to one benchmark yet, nearly 50 per cent of the Mind ETF tracks a combination of Nasdaq-100, Russell 2000 and the S&P500 indices. The stock closed at $24.92 Tuesday and has hovered around $25 since it launched in November.

Building a track record is next for the software, says Hawkins, whose investment company manages over 80 other human-managed ETFs.

Toronto-based Bristol Gate Capital Partner Inc. is also using machine learning to develop its investing model. It trawls through millions of reports and news articles with a focus on dividend-based growth, according to president Mike Capombassis.

The system is grounded in fundamentals — revenue growth, operating cash flow, free cash flow and growth, strength of the company’s balance sheet, the margins it generates, capital allocation and the industry its compete in all play a role in valuing a company.

But the technology is not at the point of being something “you set and forget”, as an asset manager has to verify its reports and make adjustments. A majority of the ETF’s holdings are U.S. companies, which Capombassis says, in terms of dividend growth, is the “deepest and broadest, by a country mile.”

Bristol Gate Concentrated U.S. Equity ETF has fallen 4.5 per cent since the stock began trading in February until markets closed on Tuesday, marginally better than its benchmark S&P500, which dropped 4.66 per cent. The fund’s biggest holdings include Broadcom Ltd., Bank of America Corp., Walt Disney Co., American Tower Corp. and Home Depot Inc.

In the future, Bristol Gate plans to diversify into global equities and mid-cap companies in the S&P 500.

Interest in the AI ETFs may surge after Blackrock, the largest asset manager in the world, with over US$6 trillion in assets, invested $108 million in a venture capital firm to fund AI startups last September.

While the possibility that AI begins replacing asset managers remains slim today,  some portfolio managers may find themselves replaced by the technology, according to Denise Davids, senior manager of Funds Research at ‎Industrial Alliance Securities. More likely, they will learn to adopt AI in their practice.

“Asset managers are using AI in increasingly creative ways, for instance, to get cleaner data, enhance client relationships, and improve back office processes,” Davids said.

Investors dazzled by AI-driven ETFs should ask themselves if the model meets their risk profile or investment style, according to Josh Blechman, director of capital markets at Exponential ETFs.

The fact that ETFs are so broadly available could mean its buyers are less primed on knowing investment risks, and retail investors may too readily buy into AI as it seems cutting-edge without fully knowing the risks. 

“The fear is that people hear buzzwords like artificial intelligence or machine learning and they invest in that, not necessarily understanding the objectives of the fund,” Blechman said.

In the U.S., Equubot’s AI Powered Equity ETF also employs data-dependent approach to parse through reams of articles and financial data to choose 70 out of a list of 6000 big companies. Its major holdings are in Radian Group, Facebook and the Nasdaq, at around a combined 8 per cent.

The system sometimes tweaks its picks daily in the U.S., CEO Chida Khatua said.

“In a sense, what we’re trying to do with it is … having millions of equity analysts going through all those companies’ information,” he said.

By contrast, Mind re-evaluates its portfolio each month, while Bristol Gate swaps out 10 to 30 per cent of its picks when it adjusts its portfolio roughly once a year.

At this point, Davids isn’t too sure what AI ETFs can deliver that’s exponentially better than flesh-and-blood managers. The strategies are still fairly new and unproven. “These types of investment are best suited toward investors that can tolerate a moderate amount of risk,” Davids said.

But Bristol Gate has already decided. With the help of human managers, its machine learning technology has started to predict dividend growth for all of its $1.1 billion of assets under management in March, Capombassis said.

soli@postmedia.com