How advisors can benefit from AI

By Mark Yamada | June 18, 2018 | Last updated on October 3, 2023
3 min read
Woman looking ahead to future while in the palm of a robotic hand - robo-advisor concept

Read Part 1 of this series from the May 2018 issue of Advisor’s Edge

Can artificial intelligence (AI) predict future market returns?

The best answer is the same one given on my kid brother’s kindergarten report card assessing his ability to play nicely with others: not yet. While AI helps perform tasks quickly and efficiently, it still seems limited by what humans can do. And how many people can successfully and consistently predict market returns?

AI makes factor and multifactor models (like value, growth and momentum) more efficient, and helps traders profitably route orders for arbitrage. But these tasks require neither intuition nor predictions.

Market predictions require analysis of historical data (a left-brain task) and assessment of behavioural factors such as herding, regret, confirmation, fear and greed (a right-brain task).

Machine learning may soon be capable of both. Neural networks that mimic the way our brains process information should be able to synthesize this information—given enough computing power, data and time—and offer insights into what triggers complacency and starts stampedes. But how can we benefit right now?

Ask the right questions

AI is probabilistic. For example, after scanning 1,000 dog images, AI may identify a photograph as having a 60% chance of being that of a dog. Understanding this allows us to ask questions in a more useful way.

Instead of asking how to maximize risk-adjusted returns—seemingly the solution to all investing problems—we could seek to improve the probability that a recent retiree won’t run out of money or that Sarah’s parents will accumulate her Harvard tuition and expenses in seven years.

Both situations pose contingent liability problems, something modern portfolio theory (MPT) can’t resolve, despite advisors’ persistence in applying MPT’s mean variance optimization to all their clients’ problems. One major shortcoming in using MPT is time. Rebalancing to the fixed asset mix suggested by an MPT-generated efficient frontier assumes the investor has time to wait for the next investment cycle and that all assumptions for expected returns, risk and correlations between asset classes won’t change (possible, but unlikely). MPT might work for the recent retiree who expects to live for another 20 to 25 years, but what is the risk of not accumulating the $86,500 for Sarah’s freshman experience seven years from now—before inflation—let alone the $346,000 for four years?

Assuming Sarah’s parents have $250,000 now and require a 6% return compounded for seven years to bank four years of tuition, AI could potentially establish an asset allocation that minimizes volatility, because any losses during the seven years could be difficult to recoup. Based on progress toward the goal, the program would reassess the probability of success and make asset class adjustments. (This is similar to a car’s GPS system, which considers traffic conditions for all potential routes.) Reinforcing each correct decision along the way to the goal encourages the algorithm to follow the correct path. This dynamic approach is the most effective resolution of the client’s problem.

Leverage industry changes

Machine learning’s biggest impact will be in data collection and analysis. The forecasted loss of 90,000 investment industry jobs globally over the next seven years (according to industry consultant Opimas) translates to about 2,250 Canadian jobs, based on GDP. PwC reports that, while technology will significantly impact 38% of all U.S. jobs, 61% of financial service jobs in the U.S. will be impacted. Administrative, compliance and portfolio management jobs are at greatest risk, the Financial Times reported earlier this year.

Investment advisors who leverage technology and AI applications will manage more clients and assets more efficiently. Delegating most administrative, compliance and money management tasks to systems that can do a better job will free advisors to develop new and existing relationships. Some advisors may mourn the loss of managing money, but investors are better served by personalized solutions to achieve goals, and technology is the best way to manage multiple goals for each client.

Portfolio managers must focus on liability-driven solutions and goals-based algorithms. Chasing returns will give way to systematically increasing the probability that investors achieve their personal goals. The sooner we realize that these aren’t the same thing, the better for our clients and profession.

As for my brother, he never did learn to play well with others—but we can learn to play well with AI.

Mark Yamada headshot

Mark Yamada

Mark Yamada is president of PÜR Investing Inc., a software development firm specializing in risk management and defined contribution pension strategies.