The Pursuit of AI-Driven Wealth Management

Understanding the application of AI to business requires an understanding of context — strategy, customers, company culture, and so forth. One application worthy of study across organizations is wealth management. A number of banks and investment firms are trying to use AI to improve that management — either to eliminate human wealth advisers altogether or, much more commonly, to augment their efforts. Our survey research suggests that while many organizations have challenges with production deployments of AI, wealth management is a clear exception.

We’ve studied wealth management strategies using AI and interviewed the analytics and AI officers who support them at several different companies. Not surprisingly, each organization has found a strategy niche in how it uses the technology to support advising clients on their investments. Their use of AI is suited to the types of clients they serve, the investment types they advocate, their overall investment philosophies, and the AI capabilities they possess. Some strategies, however, seem better (to our minds) than others. We can’t address all of the issues around “robo-advisers” — pundits love the name, but wealth management firms themselves generally dislike it — so we’ll focus here on three alternatives with radically different approaches.

Digital-Only Advice at Wealthfront

Almost all organizations that offer AI-generated robo-advice combine it — sometimes for an extra price — with human adviser consultations. Wealthfront is the exception: It has steadfastly offered only digital interactions since its founding in late 2011. Its website makes this clear, stating on its support page, “Wealthfront delivers all its services, including financial planning, investment management and personal banking, exclusively through software.”

Is that a bug or a feature? It does enable low costs: Wealthfront’s fee is 0.25% — about the same as other digital-only advisers, but certainly less than the average of just over 1% for human-only advising — and it’s free for investment accounts below $5,000. The company’s digital-only focus is consistent with fairly extensive use of AI. According to one analysis, Wealthfront uses AI in some ways that other robos don’t. The client’s answers to a risk assessment questionnaire are translated into a customized investment portfolio of cash and exchange-traded funds (ETFs) via AI. Its algorithms track client spending and saving behaviors and provide personalized recommendations to help them achieve their financial goals.

Still, although Wealthfront’s target client seems to be a strongly self-directed Silicon Valley tech millennial (it offers, for example, extensive advice on stock options), we think that this strategy is a drawback overall. Wealthfront’s growth in assets under management (AUM) has been relatively slow — now at about $16 billion. We think that the missing human adviser component is a reason for that. The company offers a lot of investment advice, but it’s mostly in the form of blog posts.

Personal Advisor Services at Vanguard

Vanguard pioneered the “get rich slow” investing approach based on low costs and index funds, and its robo-advice platform, called Personal Advisor Services (PAS), is quite consistent with that approach. It charges a relatively low fee (0.30%, or less as assets increase), invests clients’ money in mutual funds and ETFs, and generally has a conservative investing philosophy.

Consistent with its no-frills approach, PAS doesn’t do a lot with AI. Portfolios are constructed on the basis of risk assessments, but simple algebra is used to translate the questionnaires into investment percentages (equities/fixed income/domestic versus international). The primary investing goal is to finance retirements, and Monte Carlo simulations (automated tests across thousands of different potential financial outcomes over time) are used to score each client’s account for how likely they are to outlive their income. Recommendations are made for decisions like portfolio rebalancing and tax loss harvesting, but nothing is executed without adviser and client review. In short, PAS gets the investing advice job done, but with a minimum of unwarranted input or excitement from AI.

Human advice is an integral component of the program; in our experience (Tom is a client), that advice is usually oriented toward getting clients to do needed tasks (like identifying beneficiaries) and persuading them not to do stupid things (like selling stocks right after a market decline). The combination of features — and Vanguard’s very large client base, no doubt — has made PAS the giant among robo-advisers, with over $221 billion in AUM. There’s also a relatively new option with no human advice called Digital Advisor that charges only 0.15%. For the investor who wants high value at low cost, Vanguard’s offerings are hard to beat.

Next Best Action (and Increased Engagement) at Morgan Stanley

Perhaps the greatest level of AI integration with strategy is found in Morgan Stanley’s Wealth Management unit — and it’s accompanied by the greatest level of human adviser focus as well. Morgan Stanley has a big wealth practice — it’s ranked third in the world after UBS and Credit Suisse in AUM — but our sense is that both its AI and its digital strategy are far ahead of its traditional competitors’.

For over 10 years, Morgan Stanley has been working on its Next Best Action system to provide its financial advisers (FAs) with insights to present to clients. We wrote about this system, which uses machine learning to identify investments of interest and relevance to a particular client, in 2017. At the time, it was just being introduced, and the focus was on coming up with personalized investment offers. Since then, however, Morgan Stanley has also focused on the client engagement aspects of the system. The Wealth Management unit’s management team has concluded that the primary way an FA achieves success is through frequent engagement with the client — so the Next Best Action system facilitates that process. As Jeff McMillan, chief analytics officer at the company, put it in an interview, “We have a very sophisticated machine learning algorithm to identify topics of interest to clients. But in the end, financial advising is a human-based game. If all the system does is remind them that the adviser is there and looking out for them, that is often enough.”

Use of the system is voluntary, and not all FAs use it, so it is impossible to attribute AUM or other financial measures to it. But McMillan said that FAs who use it are more efficient — because coming up with relevant investing ideas is much quicker with the system — and their clients are more engaged. We think this combination of capabilities makes this a winning strategy for Morgan Stanley and other high-end wealth managers.

Is AI Alone Enough?

Other high-end wealth management companies sometimes say that AI can’t manage client portfolios that include alternative investments, such as art, commodities, or private equity. But Morgan Stanley’s McMillan said that’s not a good excuse. “There is the perception that these tools are only suitable for the ‘mass affluent’ segment and not the ‘ultra-high net worth’ space,” he said. “The argument is that the populations are too small for trustworthy recommendations. But we can drive specific opportunities based on individualized client behavior and characteristics.” He said that even if there’s not enough data for machine learning, “we can use business rules, or a test-and-control approach, to see what’s generating response.”

McMillan commented that this is not a system, but a way of doing business. He credits a cross-functional approach to managing the process, and executives who were farsighted and stuck with the idea over time. He specifically credits Andy Saperstein, the head of wealth management and now copresident of Morgan Stanley, as well as the company’s long-term chief operations officer, Jim Rosenthal, now retired.

In fact, it would appear that at all of these companies, there’s not just an AI system at work, but a new way of doing business. Each strategy is quite different from the others, and so the AI needs to be different as well. That means that you can’t evaluate an AI application in isolation from the strategy, process, and culture of the organization. Wealth management provides a useful illustration of the need to embed AI into the organizational DNA rather than add it on top.