AI could succeed, but gains may flow to capital and early adopters: Man Group CIO

[INTERVIEW]

AI could deliver on its promise of boosting productivity and still create a problem for investors if gains flow to capital faster than wages, according to a senior executive at Man Group, a London-based active investment management firm.

“The risk I think is underappreciated is not that AI fails,” said Gregory Bond, Man Group’s chief investment officer, in an exclusive written interview with the Korea JoongAng Daily and JoongAng Ilbo.

“The one thing that has clearly moved in the data is the share of income going to capital rather than to workers,” explained Bond, who also serves as Man Group’s head of the Americas and a lead portfolio manager for Man Group’s flagship multistrategy fund.

“That has been rising for 30 years, so AI would extend the trend rather than start it. It is also what happened for decades in early industrial Britain before wages caught up. If that is the path, the businesses most exposed are the ones whose customers are wage earners.”

Even years before advanced large language models were released, AI is estimated to have reduced the labor share by up to one-third of a percentage point between 2000 and 2017, according to an OECD report published in September.

Bond’s concern comes as the productivity benefits of AI have yet to fully materialize, with companies still in the midst of a massive investment cycle.

“This is the capital-spending phase. The payoff hasn’t arrived,” he added. “The AI boom differs from the dot-com bubble in that the largest technology companies have so far funded most of the build-out from their own cash flow. Yet, as with previous technology booms, there remains a lag between investment and productivity gains.”

Founded in 1783, Man Group manages roughly $253.6 billion in assets and has been actively integrating AI into its operations. After launching its internal AI research system, AlphaGPT, last November, the firm expanded its use of AI through a strategic partnership with Anthropic in February. The partnership uses Claude AI to analyze securities, generate investment ideas and uncover proprietary signals, allowing analysts to focus on company modeling, security selection and portfolio construction.

In Korea, the firm launched Global Multi-Credit with local brokerage Korea Investment & Securities in September, with the fund investing in underlying funds managed by Man Group.

Below are edited excerpts from the interview, including his views on AI investment opportunities and the use of AI in investment decision-making.

Q. Where are the AI opportunities beyond the Big Tech companies?

A. The opportunity in the next phase is in the companies that use AI, not only the ones that sell it. The thing being sold is turning into a commodity. The AI token, the unit in which model output is priced, fell from around $30 per million to below 30 cents for the same level of capability between early 2023 and late 2024. In a market like that, value moves to whoever owns something the model cannot produce for itself.

In some industries, AI makes outsourcing safer, and companies get smaller. In others, it makes large organizations easier to run, and they get bigger. So the winners are not a sector. They are the companies in each industry that build the data advantage first, though I would treat that as a theory awaiting evidence.

The practical point is that "the AI winners" is where most capital has already gone. If the framework is right, the edge is in discerning winners from losers within an industry, among ordinary companies whose economics are changing.

How is AI changing the investment process?

Within our systematic business, AI agents now work alongside our researchers, proposing hypotheses, writing code and running tests. What AI changes is the denominator.

Generate a thousand candidate ideas instead of 10, and the standard of proof each must meet has to rise, because the more you screen, the likelier the best-looking one is a fluke. The market does not produce evidence any faster because you asked it more questions. So the discipline we spend most time on is how much evidence an idea must earn before it gets client capital.

Which asset classes beyond equities suit AI-driven research?

Equities came first because they produce the most evidence. A systematic equity model ranks thousands of companies every day, and every company comes with a large volume of text. Credit is next because bond markets are moving onto electronic trading the way equities did 20 years ago, and the documentation is even denser.

Macro is the hardest case, for a reason that has nothing to do with the technology. A strategy that trades many times a day generates far more independent observations in a year. One that holds a position for six months has generated perhaps a hundred in all of modern market history, and 50 markets driven by the same global forces are not 50 separate experiments. In macro, AI will make us much faster at processing information and only slightly better at proving that a new idea is real.

What are risks of investing amid high rates and elevated valuations?

There are two assumptions that have stopped being safe. The first is that bonds protect you when shares fall. They cannot be relied on in an inflation shock. In 2022 they fell together, and portfolios that looked diversified were not when it mattered.

The second is that owning "the market" is diversified. A handful of technology companies now make up an unusually large share of the main US index, so an index investor owns a concentrated technology bet whether they meant to or not. What keeps it that way is career risk among professional investors. Own the consensus and be wrong, and you are wrong alongside everyone else. Miss the rally, and you underperform alone.

That same supply-chain position is why the same point applies here [in Korea] as to the US index. When a market's gains come largely from a single industry, an investor who owns the index owns that industry, whether they intended to or not.

So, ask what a portfolio is built to withstand, because too often, the answer is the last crisis. If shares and bonds can fall together, you need return streams that depend on neither. That is what strategies that follow trends or trade differences between securities are for, though they carry risks of their own and can lag in fast reversals.

BY JIN MIN-JI, PARK YU-MI  [[email protected]]