Robo-advice has long promised financial democratisation. But its adoption has not been the success story everyone expected. That's thanks to data security fears, mistrust, and the lack of human involvement – not to mention the rise of artificial intelligence (AI).
Marie Brière, head of investor intelligence and academic partnerships at Amundi Investment Institute, has looked extensively at robo-advice and its various barriers to entry. She says adopters of the technology have encountered significant benefits.
What Role do Robots Play in Influencing Human Financial Behaviour?
At Amundi, we introduced a robo-adviser on employee savings plans in 2017. This robot is based on human-robot interactions. It starts by questioning savers on their objectives and preferences, then offers them a portfolio recommendation. Over time, it sends out e-mail alerts if the portfolio allocation deviates from the targeted allocation, and suggests a portfolio rebalance. Individuals keep control of their investment decisions over time.
In research projects, we analysed the investment behaviour of a sample of almost 20,000 "robo-takers", compared to 60,000 non-advised clients. After using the robo-adviser, employees saved more and were more inclined to invest in equities (+9%). An important finding was most savers followed the robo-adviser's recommendations, and rebalanced their portfolios to keep allocations closer to target. These enhanced choices gave them higher risk-adjusted returns.
In particular, implementing a systematic rebalancing policy – a practice common among institutional investors but seldom used by retail investors – was highly beneficial to their performance (+2% to 3% return). Interestingly, these effects were greater for investors with smaller portfolios, who might be less likely to access traditional financial advice. This suggests automating financial advice promotes financial inclusion.
With AI, Some People Worry About Machine Bias and Algorithms Getting it Wrong? Does That Manifest in Robo-Advice?
Robo-advisers typically rely on simple algorithms. They are designing an optimal asset allocation based on a portfolio optimisation algorithm using as inputs assets' expected returns, risk, and individuals' own preferences (typically investment horizon and risk aversion). Each of these parameters can be inaccurately estimated, so that can lead to biased recommendations.
On one hand, there are substantial gains to be made from personalising asset allocation to individuals' characteristics and personal preferences. On the other, there is a risk of too much engagement, particularly when markets fall. People tend to re-profile frequently or modify their risk profile during crises. One way to reduce this source of error is to focus on the stable personal characteristics that can be precisely estimated.
Many are Worried About Data Security, or May not Trust Financial Advice Full Stop. What Should Investors be Aware of?
We know trust is a key driver of financial decisions. Trustful investors are significantly more likely to invest in the stock market or follow financial advice. But trust is also a key component of successful robo-advice, and determines an individual's willingness to follow the robo-adviser's recommendations.
Survey evidence shows there is a general lack of trust in algorithms. While most people seem to trust their general environment and technology, AI is not yet trusted.
There are different ways to build trust, however. The first is to give people some control over the algorithm. In scientific experiments, it's been shown that professional forecasters are more likely to adopt an imperfect algorithm when they can modify its forecasts, even if they are severely restricted in doing so. One way to build trust is to let humans and robots interact, with the robot proposing an advice and the human being the ultimate decision maker. Another way to build trust is through explainability: i.e., offering clients an explanation of the recommendation, even if it is based on a complex model.
How Can Robo-Advice Businesses Address These Concerns and Increase Adoption of Their Services?
Incorporating more complex AI into robo-advice will be challenging. But recent advances in so-called XAI (explainable artificial intelligence) could be useful.
"Explainability" refers to the possibility of explaining a recommendation even if it is based on a complicated model. But we shouldn’t look for full transparency of the algorithm underlying the robo-adviser. It would be more effective to disclose, for example, which economic scenarios may cause the algorithm to perform less accurately, or informing clients about the overall limitations of the algorithm itself.
Another potentially interesting development would be to strengthen the interactions with clients. Some robo-advisers send alerts when a client's portfolio deviates significantly from its target asset allocation. Such alerts could also be seen as an opportunity to interact with the client. For example, they could be used to explain why a deviation occurred (market movements, change in personal characteristics, etc.) or why rebalancing is recommended. As such they too can become a tool for financial education.
