HAL

AI: Buzzword or Bonanza?

Overhyped or not, its underlying technologies are going to change, in fundamental ways, the financial advice industry.

By Patrick Beaudan

It’s true that AI, the shortened form of artificial intelligence, is the industry’s favorite buzzword, but there’s a reason for that: AI holds keys to making financial advice simultaneously more scalable and more individualized. What a powerful combination! Whether you call it machine learning, algorithmic applications or Baby HAL, AI is already helping the financial world operate more efficiently in a multitude of ways, from trading stocks to picking out potentially fraudulent credit-card transactions.

Today, a lot of algorithm developers are working on portfolio construction.  Ideally, AI would make portfolios more responsive and dynamic in real time, and more effective at managing risk. This is a noble task—but also one that will require a tremendous amount of data and experimentation over time in order to prove itself.

We believe, however, that the AI application that would be most beneficial for financial advisors and investors would help with behavior, rather than trading. Most people are overburdened and distracted. It’s hard to concentrate on the big picture. Financial behavior is always a fight between strategy and impulse—and the less time you have, the easier it becomes to fall back on impulse because you don't have time, or are too stressed, to go about the research and critical thinking needed to evaluate and update your strategy. This is most common among clients, but occasionally true of advisors as well.

That’s why we believe that both advisors and investors deserve an “early warning system” that could detect the signs leading to behavior change. Advisors, for example, could make good use of an application that notified them when clients exhibited behaviors signaling that they were thinking about leaving. Although the final decision to leave a practice might be impulsive, there is often activity that points to diminishing engagement. A smart algorithm could recognize toxic combinations of diminishing client clicks, market data, account information, and research on social media (assuming that the advisor has access to the client’s posts).  The various data bits would have to be placed on a scale of relative importance so the algorithm would figure when to activate the alarm. It’s a delicate balance—clients will be annoyed if their advisor calls them too often, just as they’d be disappointed, and perhaps become disaffected, if their advisor doesn’t call them enough.

Investors need a different early warning system—one that pings them if their portfolio needs attention. Of course, this should be the advisor’s job but sometimes investors want to be told first.  And sometimes it’s the advisor who’s distracted! The algorithm could look at whether a portfolio has drifted out of its proper allocation, or is not keeping up with market movements as it should. These outcomes are a little simpler to produce, because the data they need is relatively straightforward. Investors then have the option to reach out to their advisors or take action themselves—whether that means deciding to wait, engage in some research, or reposition the portfolio.

I Need Data. Lots of Data

Even the simplest AI applications require a tremendous amount of data to be effective. An algorithm – the basic unit of artificial intelligence – is basically a mathematical rule. “Machine learning,” the quality that makes the algorithm intelligent, means that data fed to the algorithm also refines it, making it more precise. To use a simple analogy, think about a recipe for pancakes. It tells you how much flour, how many teaspoons of salt and baking powder—but each time you make pancakes, you tweak the recipe, perhaps adding nuts or chocolate chips, for example, to get the fluffiest, most flavorful pancake you can make. 

Financial algorithms require a lot more data than you’d think—some proprietary, some publicly available. The more data, the more ability to design algorithms that get the right results. How much data can you get on a person? You can buy data on social flows, demographics, and custodial data. These allow a programmer to triangulate and build a profile. For instance: The demographics of people who live in your zip code, of the same gender and age group, with the same number of children. But all this data has been anonymized.

But an app creator can only collect data on you personally with permission. Let’s say you friend your clients on Facebook. Expecting a grandchild? Going on vacation? It’s there. An app developer can go down the complexity tree and figure out what affects clients’ financial behavior most profoundly. Then the developer has to store the data securely and compile it with an algorithm.

If the application and its underlying algorithms are well crafted, this data has the potential to improve the relationship between advisors and clients. To make advisors proactive—and even more important, to make their activity timely and pertinent. “Why do I need to call my client today?” the advisor might wonder. The program will say why. Armed with AI-generated insights, an advisor can prepare a plan that is more meaningful, better attuned to a client’s needs, whether expressed in conversation or by behavior. That is why I believe that AI will make financial services both more scalable—the machine finds who you need to call faster and more efficiently than you ever could—and more individualized. But you still have to talk to the client. Your ability to create an empathetic connection with your clients is irreplaceable.

Less Than Human 

More important, AI will not replace us. It doesn’t know how to talk a worried client out of going to cash, or how to tell a client’s adult children that the client is showing signs of dementia. Financial relationships have deeply human elements that AI won’t be able to address for many years, if ever. 

But AI will be a mainstream element in financial services soon enough. We need to get beyond the fear—and the hype—and start working with code. A lot of valuable breakthroughs will be coming soon. 

Even the most intelligent algorithms are no match for humanity in the world of finance. Although IBM’s Deep Blue defeated world champion Gary Kasparov at chess as long ago as 1997, the program could play no other game. Google’s DeepMind, a highly advanced artificial intelligence program, is still directed to specific tasks (at least so far). It has no individual agency, no will. In finance, the algorithm that trades stocks may know nothing of bonds or real estate.

The reality is, you shouldn’t give too much credit to what a machine can do. Somebody has to think about all the potential scenarios, take in the responses from clients, then figure out how to create the algorithm, feed it enough relevant data to make better choices, then refine it again and again. 

Is a client-behavior early warning system conceptually possible? Yes. Will the app be fully operable from Day One? Not in a sophisticated way. At first, it will likely be rough; then with much feedback over time, the industry will get something reasonably powerful. That will be a good day for advisors, and a great day for clients.

Patrick Beaudan is CEO of Emotomy, an investment platform and robo solution based in Tiburon, Calif., and CEO of Belvedere Advisors LLC.

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish