If the use of artificial intelligence is not yet ubiquitous in the wealth management industry, the discussion of it certainly is.
To help break through some of the hype and hand-wringing, we sat down with William Trout, senior analyst at research, advisory and consulting firm Celent to discuss how AI is being used in the financial services industry – including its impact on flesh-and-blood financial advisors. The good news: work where “empathy, judgment or humor” are required will still call for the human touch.
Q: What are some ways firms use AI to find investment opportunities?
WILLIAM TROUT: Hedge funds and so-called quant investors have long used AI to uncover trends and build upon what they see as their data advantage. They use data mining and pattern recognition to help them flag or highlight trends that the human investor would not otherwise recognize or consider. An increasingly visible use case in this context is for security selection. BlackRock’s new Advantage funds are entirely machine-driven. In fact, several dozen human portfolio managers left the firm as part of the fund launch.
AI is particularly powerful in cases where there is information asymmetry, or a gap in knowledge between buyer and seller. This gap has historically been visible around asset classes like commodities or around certain types of stocks, like those of retailers. Satellites that are able to peer down on crop growth in Iowa or the Wal-Mart parking lot can deliver data that can be used to extrapolate prices or profit margins, for example. These advantages are much more difficult to realize in cases where information is closely held or unavailable, as in the case of private equity.
Q: Does AI obviate the need for human perspective in asset selection?
WT: AI relies on the use of massive computational firepower to tease out trends and identify opportunities for rapid investment decision making and execution. Bigger and faster is not always better, however. Hedge fund AI models based on human thinking may end up emulating the worst characteristics of human investors. Part of the problem is the tendency of machines to “overfit;” that is, to project or model outcomes based on the information known at the time, which is by definition incomplete.
Here we touch on the intrinsic limitations of machine-based technology. AI derives insights from sifting historical data; it is at core backward-looking. This inhibits its ability to draw conclusions about future developments or even identify patterns that might cause historical relationships to break down. Human intelligence is needed to moderate or correct the assessments of the machine, as well as to explain and communicate decision. So-called “deep learning” in the form of neural network technology has attempted to address the gap between human intelligence and AI. For now, the person holds the upper hand, in that AI functions best as a tool for amplifying the capabilities of the human being. This is true particularly in the complex, service-driven culture of wealth management.
Q: For financial advisors, what about incorporating AI into customer relationship management (CRM) tools?
WT: Most, if not all, wealth management firms already use customer relationship management (CRM) software to keep track of client data and interactions. CRM providers like NexJ and Salesforce have invested heavily in building out their AI capabilities. Salesforce Financial Services Cloud uses its Einstein artificial intelligence agent to re-score bank-provided analytics, such as on leads and activity cycles, to anticipate and respond to individual customer behaviors. NexJ has repositioned its Customer Data Mart as CDAi (Customer Data Analytics and Intelligence Platform) to underscore the degree to which machine learning helps advisers move from data to insights to action.
Q: What questions should firms consider before they try to implement AI in their workflow?
WT: Cloud-based collaboration tools like Slack and Symphony that integrate with modern CRM technology offer a fast and low-cost means of generating insights. In the long term, however, firms will need to develop technology road maps that support the more robust treatment of data and closer integration of that data to the portals and dashboards used by business analysts and advisors.
Firms need to get their data houses in order before undertaking transformational CRM-led AI initiatives. The problem is that the IT systems of banks and other large wealth management organizations are still defined by iterative development cultures borne of short-term fixes and corporate acquisitions. The result has been spaghetti systems and isolated, single-use, case data marts. Enterprise data warehouses and data lakes were set up as a way to centralize and render accessible the information housed in these silos. Unfortunately, an explosion in the volume and sourcing of data has created an even bigger mess: the data swamp. Today data is sourced from inside and outside the organization, via APIs and via new channels, including mobile devices, social media and the Internet of Things.
Q: In the wealth management industry, will AI impact client communication?
TROUT: Wall Street research (for example, on a stock or company) and earnings coverage have been among the early use cases for AI married to natural language generation technology. In fact, any communications that relies on a structured set of data will be more efficiently processed using natural language generation. Think of a baseball box score in which the inputs vary but the way the information is set out remains the same. The speed and accuracy at which the trained machine can convert such data into prose cannot be matched by the human being. Obviously, cases where empathy, judgment or humor (to name a few examples) and similar qualities are required still call for the human pen.
Machines are good at crunching massive data sets. They are less good at deducing context or parsing nuance. Some level of human oversight will generally be required, even in cases where a premium is placed upon timeliness, such as updates delivered over the Internet. The case of Tay, Microsoft’s hate-spewing chatbot, underscored for many the need for closer human supervision, given the extent to which AI driven tools can be gamed or even go rogue. It’s a classic case of measure twice, cut once!
April Rudin is the founder of The Rudin Group, a financial services consultancy.