Every day, a new article proclaims that artificial intelligence will transform wealth management. While some tout AI-based “solutions” that are highly suspect, most of these articles reduce to vague explanations of how AI will enhance robo-advisors and how generative AI like ChatGPT will improve research and client service.
I do not doubt that robo-advisors will continue to improve and that generative AI will lead to efficiencies and cost reductions (although the potential utility of large language models like ChatGPT is diminishing with each new research paper and corporate ban, increasingly limiting its application to mundane tasks). But both will be quickly commoditized, and neither will prove truly transformative.
There is a transformative technology that will provide wealth management firms — especially wirehouses and large independent broker-dealers and RIAs — with a sustainable competitive advantage, a strong return on investment, and better investment outcomes for financial advisors and clients: the recommendation engine (RE).
An RE uses advanced machine learning and statistical modeling to anticipate customers’ needs and wants based on a business’s unique historical and behavioral data. It produces recommendations based on a combination of a customer’s past behaviors and history, a product’s consumer ranking, and the behaviors and history of similar products.
Readers might not be familiar with REs, but as Jan Teichmann, principal data scientist at Trainline, observes, REs “are everywhere and for many online platforms their recommendation engines are the actual business. That’s what made Amazon big: they were very good at recommending you which books to read.”
Amazon’s RE accounts for 35% of its revenues. At Netflix, its RE is so central to the business that its data scientists and designers literally reengineered the company’s user experience around the motto, “Everything is a recommendation.”
Michael Schrage, a visiting scholar at the MIT Initiative on the Digital Economy, explains the transformative power of RE: “These ingenious mechanisms relentlessly convert data into relevant, diverse, novel, and even serendipitous options. They learn from the choices people make, explore, and ignore. That means they are the irresistible — and inevitable — future of innovative advice.”
REs may sound like robo-advisors and other analytical financial planning tools that also make recommendations. However, these other tools are generally based on decades-old modern portfolio theory to determine an investor’s risk-return preferences and automatically (in the case of robo-advisors) select investments and rebalance client portfolios. Their output is, therefore, bound by human knowledge.
By comparison, REs’ power is based on a type of machine learning called reinforcement learning. Because of its learning architecture, each of an RE’s recommendations improves the accuracy of its subsequent suggestions.
The key to a successful RE is data — lots of data. Fortunately, legacy wirehouses, large independent broker-dealers, and RIAs are veritable digital data warehouses. Standardized forms capture vast amounts of client-specific demographic, geographic, financial, and psychographic data, and internal systems capture client behavior through trading activity.
In the same way that Netflix’s RE uses a subscriber’s viewing history and taste preferences (as well as those of similar customers) to provide “highly personalized recommendations,” a properly trained wealth management RE would use its client data and the data of comparable investors to regularly make personalized investment recommendations.
For example, based on a client’s risk-return profile and investment objectives, changes in the market environment, and successful changes made to similar clients’ portfolios, a wealth management RE might suggest a tactical shift in a client’s portfolio away from equities and into fixed income.
The benefits of an investment RE would be manifold: Financial advisors would have a tool that empowered them to consistently engage with clients and allowed them to make better recommendations than could be made with traditional analytical tools (e.g., robo-advisors). Better recommendations would mean better performance, which would lead to higher satisfaction and higher retention of both advisors and clients. The recommendations would also lead to more frequent (and accurate) trading activity, which — given wirehouses’ and broker-dealers’ payment-for-order-flow revenue model — would increase a firm’s revenues.
Yet the path to a wealth management RE is fraught with challenges. Most fundamentally, adopting an RE requires the wealth manager and its advisors to move away from the inculcated view that investing is essentially a human activity and accept that advanced machine learning like reinforcement learning can be used to make better investment decisions than human intelligence-based tools. This attitudinal shift must start in the C-suite and ideally with the board of directors, with an unambiguous, firmwide declaration affirming that adopting an RE is a move to empower, not replace, advisors. In fact, an RE would allow an advisor to spend more time on critical financial planning matters like behavioral coaching and customized wealth, retirement, estate, and tax planning.
Other challenges are technical. Wealth management firms have vast amounts of usable data to feed an RE, but their data is typically siloed across the organization. Second, building an RE is a complex, expensive project that requires a dedicated research-and-development infrastructure and specialized data science talent that is difficult to attract and retain. And even with the right data and talent, designing, developing, and commercially deploying an RE is a heavy technical lift. Success is not guaranteed for any organization, especially in the wealth management industry, where companies report that “50% or less of their emerging technology projects were successful.”
But if a wealth management firm wants to achieve a sustainable competitive advantage, it needs to think beyond robo-advisors and ChatGPT and build its own RE — despite the challenges.
In failing to take this approach, wealth managers risk being overtaken by a direct competitor or an adjacent investment firm (think BlackRock and Aladdin) or being disrupted by an outside actor (think Amazon, Alphabet, Apple, etc.) that’s looking for a significant diversifying revenue stream and already has all the required ingredients to create an investment RE.
Better to think like Amazon than be eaten by Amazon.
Angelo Calvello, PhD, is the co-founder of Rosetta Analytics, an investment firm that uses deep learning and deep reinforcement learning to build and manage investment strategies.