“Technology is changing too quickly for us,” sighed the head of IT at a US-based family office, while taking a break from SS&C Technologies’ annual client conference in September at Las Vegas’ Wynn casino. In the 100-degree heat of Nevada in late summer, the topic of conversation throughout the entire event tended toward artificial intelligence, machine learning and other emerging technologies, but for some, they may as well have been talking about fairy dust.

“We’ve only just got [SS&C’s back-office platform] Geneva,” the IT head continued, while stubbing a hand-rolled cigarette out amid plumes of smoke from electronic vaping devices. “I don’t even know how I’d begin getting the budget to hire mathletes and machine-learning specialists.”

It’s a common problem for many on the smaller end of the buy side, who do not have the staff, the budget, or even the inclination to pursue emerging technologies like AI. While an arms race continues to heat up between the largest or the most tech-savvy global firms as to which datasets can be analyzed and inform a trading strategy before they go stale, many sub-$10 billion asset managers are content to sit back and let them fight it out—mostly because they have little choice in the matter, even as they admit subsets of AI will become standard in the not-too-distant future.

AI as a Solution, Not a Problem

To understand why AI’s various subsets—machine learning, natural-language generation and processing, robotic process automation (RPA) and others—are becoming so important to the buy side necessitates a knowledge of the current environment. Much has been written about diminishing returns from traditional strategies, and the appetites of investors switching from active management to the passive profits of exchange-traded funds and index trackers, for instance. 

But there are also other worrying signs of stagnation—a generally flat number of family offices and smaller funds or asset-management firms opening year-on-year, without any significant growth, is one. The consolidation in the providers of technology to buy-side firms, as seen with State Street’s acquisition of Charles River, and SS&C’s subsequent purchase of Eze Software, which is due to close in October, is another.

The use of alternative data, in particular, has emerged as a sweetheart topic for many shops. The idea is that if the vast morass of data out there on any number of topics can be wrangled and organized into some form of coherent structure, patterns can be discerned that can lead to alpha, for those with keen eyes. For instance, while the example of counting cars in mall parking lots as a gauge of, say, a commercial real estate investment trust’s performance is well-known, enterprising firms are also tying together seemingly disparate elements, like aluminium stockpiles at factories that serve in the supply chain for toy companies, in order to predict box-office performances of major films, using the level of merchandizing as a proxy.

All of this requires AI. Humans simply cannot analyze all of that information quickly enough to make a difference. That’s all well and good for the small tech-powered buy-side firm in Hoboken, with 20 PhDs on staff, or the really sophisticated heavy-hitters. But even for those with the nous, keeping up with the capabilities can be an issue. Outside of alternative data, however, the justification for engaging in AI research and the return on investment can sometimes be more difficult to ascertain.

“Over the last four years, we spent a good bit of time addressing issues related to optimizing scaling within our operations,” says David Sharpe, managing director and director of operations for Fortress Investment Group’s credit business. “The challenge we have at Fortress is that we’re not a shop that’s going to have 3,000 trades per day, so identifying large-scale operational efficiencies is tough for us, in that a lot of what we do is fairly bespoke. We went to a lot of conferences, a lot of panels, and walked away thinking it would be great to use RPA but we’re not sure how relevant it would be. It took some of that frustration to think about how we could identify opportunities within operations that would be more suited.”

In the past six to eight months, Sharpe says, the alternative asset manager has put a number of components into production. Most early work, as with many other firms, had focused on RPA—a means of automating repetitive, rules-based processes that have usually been performed by humans in the past. From there it shifted into more advanced forms of AI, such as machine learning.

It’s a starting point for most firms, most of which say that AI tends to be the natural solution to many problems for their clients.

“Our clients are financial advisers and are not usually [early adopters of technology]. I don’t get a lot of questions about AI, aside from asking what it is,” says Dani Fava, director of institutional product strategy and development at TD Ameritrade. “An advisor will call up and ask how they can get a client to stop calling and asking them about the status of their paperwork—we have chat bots, we have RPA behind the scenes processing these workflows. The questions we’re getting from clients are ones that lead us to AI solutions.”

RPA is relatively simple to put in place, even if banks such as Societe Generale and others have generally found it to be flawed for their purposes. It is, after all, essentially a glorified macro—what truly differentiates machine learning is the ability for that macro to teach itself how to do something that it wasn’t programmed for. That’s where the real value—and the cost, as well as unforeseen issues—comes into play.

Resource Management

“This is a business of returns,” remarks a head of another family office based in Florida. “I can’t afford to waste money on experimenting—show me that it will add to my bottom line, or that I have to do it, and I’ll invest in it. We leave the tinkering to people who do that for a living.”

While defensive, this reaction also sums up many of the issues at play with smaller firms—AI is an unproven technology, at its core, and there are more than enough demands on firms already. This is true even for the sell side, which has traditionally led in emerging-technology development. A recent study by Gresham Technologies and the Financial Industry Markets Association, for instance, found that while the respondents were exploring the use of AI heavily, data management projects were seen as a higher priority by the vast majority—around 89 percent.

AI also doesn’t happen in a vacuum, where a firm can decide one day to download a developer kit from Google and be up and running. It requires organized and well-managed data structures, as well as a trifecta of talent which can often be difficult to attract.

“Essentially, you have the three core people involved—one who understands the business problem, and that’s the most important person,” says Steven Miyao, president, AI and analytics, research and compliance solutions at SS&C Technologies. “That person needs to understand the industry, to have been in it, and exactly what the problem is. Then you need the actual data scientists, the people who can write Python and handle the advanced mathematics. But you also need someone in the middle who can translate between the business person and the data scientist.”

So what are the solutions for cash-strapped—or, at least, cash-conscious—firms that are unwilling or unable to invest in AI? One approach is, of course, to leave it to systems providers, many of which have dedicated research and development arms, labs, innovation hubs and other such initiatives in place already. Eventually that technology will filter down into the products used by firms, although as Fortress’ Sharpe says, that doesn’t really help those businesses whose requirements tend toward the bespoke.

“The beauty is that, today, there are a lot of service providers out there and you don’t have to build stuff for yourself anymore,” says SS&C’s Miyao. “There are things that you can and should, but working with service providers, you can outsource a lot of that. This enables smaller companies to be able to do things that probably, a few years ago, they had to build for themselves.”

David Easthope, a senior vice president at research firm Celent, agrees, saying that third-party providers are “all over this.”

“Machine learning is something they really want to do. So you can rely on your third parties and systems integrators to do this,” he adds.

This is likely to be the approach taken by many firms, even for alternative data analysis—the head of the family office asks Waters if we “truly believe FactSet and Bloomberg won’t be giving me this stuff in a few years”—but another route is the appointment of an advisory board, particularly if a firm is sensitive to compliance concerns. TD Ameritrade’s Fava, for instance, recalls one of the earliest of the firm’s experiments in AI was when the emerging technology division synthesized the CEO’s voice. Once the compliance department heard about it, she says, they came down on them like “hellfire.”

Others are using such mechanisms to sound out the market before diving in. Last year, Andrew Powers, head of IT at Florida-headquartered Polen Capital, told Waters that the firm had done just that, to avoid going full-bore on technologies that may end up being irrelevant.

“We’ll be ready to adopt it, but it has to become accepted by the industry, the regulators, and everybody like that first,” he said, adding that he wasn’t certain AI was as applicable to his firm, which has a very defined investment style, as it was to those that engaged in factor-based investing.

As the head of IT at the US family office said at the beginning, trying to keep pace with the head of the pack is not always the wisest move. For those unable to compete on this playing field, slow and steady might yet win the race—even if it takes a while to get there.

“Technology always seems to move both faster and slower than you think,” said Powers. “Sometimes things happen and you’re left wondering how they happened so fast, but when you’re actually going through it, it seems to take a really long time.” 

However, it is going to be tough for those who do not invest, or who actively avoid engaging, to continue to compete. Fortress’s Sharpe says the firm is looking at “trade capture, confirmation and settlement across all asset classes, and how we can use machine learning to facilitate those for us,” and predicts that in the alternatives space, many firms will be using RPA as a spring board to then go into machine learning. TD Ameritrade’s Fava says that AI will eventually shape the future of asset allocation, which should set alarm bells ringing among smaller shops dependent on a few key accounts.

Ultimately, however, it comes down to a decision on whether AI is a must-have, or a nice-to-have. That, most say, depends on each firm’s individual goals, and while Fava describes it as “critical” and “unavoidable” in the future, others suggest there is still an element of choice.

“It depends on your motivation,” says Fortress’ Sharpe. “If the internal motivation is cost avoidance then it’s probably a must-have. The sell side is pushing it, but there’s a bit of a gray line between a nice-to-have and a must-have.”