<!-- Swiftype Variables -->

Special report: Artificial intelligence in money management

Capabilities key to firms’ future success

'Arms race' underway to make better sense of data

The growing ranks of money managers looking to harness artificial intelligence, machine learning and big data capabilities in investment operations will need focused, strong leadership to gain a competitive advantage by doing so, gatekeepers say.

Over the past two years, the growth in the number of firms “seriously dedicating time” to exploring those capabilities has been significant, said Richard Dell, a London-based partner with Mercer and head of the firm's global equity manager research team.

“We would hope the managers we speak to ... understand the way the industry's evolving, the new ways of accessing the data — and are trying to do something about it, trying to stay ahead of the game,” he said.

Those who sit out the “incremental arms race” underway now to make better sense of that data risk being “left behind,” Mr. Dell said.

Left behind or worse, some money managers argue.

Incorporating artificial intelligence and machine learning will be more than “additive” for broad-based money managers, said Anthony Lawler, co-head of GAM Holding AG's London-based quant affiliate, GAM Systematic. “If you don't have it, you may well go out of business,” he said.

GAM Systematic, which uses machine learning, was managing $4.7 
billion of GAM Holding's $163 billion in assets under management as of Dec. 31.

While it's still early days for the broader asset management industry's embrace of big data, the pressure to do so should become greater over time, analysts said.

With the exponential growth of data being generated around the globe, the advantages a manager with cutting edge capabilities in analyzing big data will enjoy — in terms of “seeing reality more clearly” — should grow, said Tony Ho, Hong Kong-based co-founder and managing director of Sandalwood Advisors, an alternative data platform focused on Chinese consumer spending trends.

The likelihood that high-powered algorithms will play an ever-greater role in helping managers make sense of that data poses challenges for gatekeepers serving institutional clients — both those newly focusing on AI and machine learning as well as those with longer histories covering the sector.

Parsing results

The rise of AI and machine learning will require investment consultants to upgrade their own abilities in quantifying and authenticating how those cutting-edge capabilities are contributing to a manager's investment results, said Alan Kosan, senior vice president, head of alpha investment research, responsible for fundamental/quantitative manager research, with Darien, Conn.-based investment consultant Segal Marco Advisors.

Mr. Kosan said Segal Marco has just begun the learning process to position its manager research team to better “parse through ... performance results” to determine what to attribute to AI, relative to the human component.

Likewise, consultants that can point to longer histories in focusing on AI and machine-learning strategies describe a state of considerable flux.

NEPC LLC has a detailed process that allows it to “vet and invest” in the “deeply quantitative strategies” it has focused on for more than a decade and, for over five years, what has become known as the AI sector, said Timothy Bruce, director of traditional research with the Boston-based investment consultant.

But “we're always researching new ways to evaluate” that quickly evolving space, “so I would expect our process and investment criteria to evolve,” he added.

Cambridge Associates LLC is using traditional qualitative and quantitative measures to improve its ability to evaluate managers' success in employing AI and machine learning, but the “holy grail” for the Boston-based investment consultant is development of a machine-learning tool that would detect “manager investment skill when it is actually present and eliminate false positives,” said Adam Duncan, managing director and head of portfolio modeling and analytics teams at Boston-based Cambridge Associates.

“We're looking for ways to build skill-detection algorithms that accurately identify the traits associated with successful money managers who possess repeatable skill that produces outperformance,” he said.

Consulting veterans predict it could be awhile before the industry reaches more definitive conclusions about AI and machine learning.

There'll be a growing universe of “dual” products going forward that use AI as a component, and the coming 10 years will be a “testing period” for how effective those strategies prove to be, predicted Fraser Murray, director of research with Melbourne, Australia-based investment consultant Frontier Advisors Pty Ltd.

Mercer's Mr. Dell posited a broader takeup — “ultimately everybody will be moving to use whatever technology gives them the best chance of getting good outcomes” — and an even longer testing period. “We're probably looking at decades … before we know how successful some of these processes are,” he said.

Big managers stepping in

One clear trend: Interest in AI and machine learning is extending to big, traditional money managers now, moving beyond the hedge funds and quant firms that have been best placed to exploit the signals big data analytics offers on mostly short-term and capacity constrained investments.

Gatekeepers concede the evolving importance of big data analytics could prove challenging for those traditional managers.

“Asset management firms, if they're successful, have a very large and profitable business, and they're run in a certain way ... and to make changes to that takes time,” said Mercer's Mr. Dell. “It's not easy. It takes strong leadership.”

Among asset management companies whose reputations were built on the skills of their portfolio managers and analysts, Fidelity Investments has been one of the more prominent to embrace the big data challenge in recent years.

The Boston-based firm has ramped up its use of machine learning and embedded data scientists within its investment teams, for fixed income as well as equities.

Fixed income — an over-the-counter market with less readily accessible information than equities — is particularly deep and prime for data wrangling to help portfolio managers find alpha sources, said Eric Golden, a managing director and fixed-income portfolio manager with the firm.

New machine-learning algorithms are paving the way for a much deeper analysis of Fidelity's huge proprietary database containing information about millions of bonds. “The fixed-income game was about turning over stones,” Mr. Golden said. Going forward, analysts and portfolio managers will need computer science skills to write the algorithms they'll need to analyze bond data, he predicted.

Other firms are joining the fray. Traditional asset management firms now account for 10% of the roughly 50 clients accessing Sandalwood's China-focused data sets, noted Mr. Ho. Three years ago, the firm's client list was made up exclusively of hedge funds.

Pitfalls abound

Industry veterans warn the road to machine learning becoming a ubiquitous money management tool could prove a rocky one, with many firms wading into the field now having little appreciation of the costs and challenges involved.

“You can't just hire a Ph.D. or two and hope that it's going to work out,” said Eric Nierenberg, chief strategy officer with the $67 billion Massachusetts Pension Reserves Management Board, Boston.

Money managers might have unrealistic expectations about their ability to leverage machine learning and “robotic automation” to produce alpha, said Cambridge Associates' Mr. Duncan. “Throwing money at the problem,” or a bunch of Ph.D.s in a room, won't necessarily produce better outcomes, he said.

“It's not something you can build overnight,” said Philippe Jordan, president of CFM International, a Paris-based quant shop with $11 billion in assets under management. CFM's 228 employees include 60 researchers — mostly with doctorates in physics or math — and 100 data scientists.

“You need literally years (of) trial and error and time” to turn rigorous Ph.D.s into rigorous market people designing models, said Mr. Jordan. The “depth of knowledge that you need in terms of scientific experiment … to be able to tell the difference between gold and fool's gold in analyzing data, that comes with a lot of mistakes,” he added.

While conceding that terms such as AI mean different things to different people, most market veterans predict the impact of wider adoption of machine learning on the asset management industry will prove more evolutionary than transformational.

Making the leap

The wild card in those assessments is whether AI can make the giant leap from its current strength in making sense of impossibly large sets of data to eventually one-up humans in making longer-term predictions about financial markets.

“Most ... signals have an incredibly short half-life — more a matter of how much money can be put to work in four weeks than four years,” said NEPC's Mr. Bruce. That fact has narrowed the pool of clients open to embracing AI-focused strategies, he said.

Jason Hsu, chairman and chief investment officer of Hong Kong-based Asia quantitative equity boutique Rayliant Global Advisors, said he's skeptical the day will ever come when computers can best humans in managing portfolios over the longer-term, as opposed to helping them do so.

Applying hard science to a social science never works, said Mr. Hsu. All of the market-related data available to feed into an AI engine are path-dependent, making the conclusions it spits out exercises in data mining, he added.

Some consultants, while agreeing traditional asset management jobs are mostly safe for now, won't rule out the possibility things could change.

“Certainly, the technology available at the moment doesn't have the capabilities to draw the (long-term) insights you need,” Mercer's Mr. Dell said, but “potentially who knows, at some point in the future, that's where we get to.”

“No one really knows today how far this could go,” agreed Segal Marco's Mr. Kosan, who predicted it won't be long before AI begins “to penetrate the investment management industry,” with the copious data it generates making it a “great laboratory, a great environment” for AI to flourish.

That backdrop prompted Segal Marco's manager research team to start asking managers to detail what they're thinking and doing about artificial intelligence now, said Mr. Kosan.

Other observers predict financial markets should prove relatively immune to the disruptive potential of artificial intelligence.

There's simply more inherent uncertainty when it comes to financial markets, compared to other parts of the economy, said MassPRIM's Mr. Nierenberg. AI and machine learning, broadly defined, are already identifying short-term trading patterns managers can exploit, but there's no clear evidence yet that they're able to identify new patterns that will persist over time, he said.

“I'm not going to say AI is never going to happen (but) until we see more of a sustained record of success, it's not super-appealing to me as an investor,” said Mr. Nierenberg, who also oversees MassPRIM's hedge fund program.