Artificial intelligence is changing the world of retirement planning. By using improved datasets and algorithms to efficiently deliver solutions tailored to people’s needs, AI can help them save, invest and retire better. One of the hottest trends to emerge in this area in recent years is the use of robo-advisors. These are software programs that use the data supplied by clients to create and automatically manage their investment portfolios. They’re gaining in popularity, but are they better than human advisors?

“Robo-advisors are a potential solution to the complexities of financial decision-making,” particularly in retirement planning, said Jill E. Fisch, law professor at the University of Pennsylvania. “But at the same time, there’s a lot we don’t know about robo-advisors — exactly how they work and how effective a solution they’re going to be.” She and other experts from Wharton and elsewhere spoke at a conference hosted by the Pension Research Council titled “The Disruptive Impact of FinTech on Retirement Systems.”

Robo-advisors, or robos, are online services that use algorithms to automatically perform many investment tasks done by a human financial advisor. Initially offered by startups, robos are now part of the suite of services offered by major financial institutions such as Vanguard, Schwab and Fidelity. Since they are less expensive than a human advisor, they democratize access to financial advice. Robos can take on customers with little savings since adding one more person wouldn’t cost much more.

Signing up starts with consumers filling out detailed questionnaires online about their financial goals, risk tolerance and investment timeframes. Robos take the information and use computer algorithms to come up with an asset allocation that fits the customer’s needs. Once the portfolio is created, robos also manage it, doing things like rebalancing the portfolio, executing trades, performing tax-loss harvesting and other actions.

“This is really something that’s welcomed by the vast majority of retail investors who find themselves inadvertently … tasked with the responsibility of managing their financial well-being,” Fisch said. “People don’t want to do this and they don’t want somebody to give them advice on how to do this. They want somebody to do this for them. That’s the space … robo-advisors are going into.”

“We don’t know in a complete cycle what these [robos] are going to deliver…. The real issue is, does it improve outcomes?”–Brett Hammond

Robo Fees and Returns

Robos came on the scene about a decade ago, and two early startups were Wealthfront and Betterment. Today, there are dozens of robos in the market, Fisch said. There are pure robo services, as well as those that offer the option of talking to a human advisor, with or without an extra fee. Since they’re automated, robos can more easily avoid conflicts of interest that could beset a human advisor, who might push investments that pay the highest commissions.

Robo fees can range from zero — if the investor has less than $10,000 to invest — to as high as 0.89% of assets under $1 million in some cases, said Brett Hammond, research leader of Capital Group. But 0.25% to 0.30% of assets is more typical, he added. (The fee is on top of the cost of the investment itself.)

As for performance, it’s a mixed bag with some robos doing better than others, Hammond said. The big question is how they will do in the long run, especially during a big market crash, since they don’t have an extended track record yet. “We don’t know in a complete cycle what these [robos] are going to deliver,” he said. “The real issue is, does it improve outcomes?”

Kent Smetters, Wharton professor of business economics and public policy, has his doubts. Robos put investors into ETF portfolios for “professional fanciness” but they are “not really a better value” than three Vanguard index funds invested in the U.S. total stock market, international securities and bonds, which are more tax efficient and costs even less, he said.

Robos also overly rely on risk tolerance levels they assess from questionnaires. But academic studies show this metric is not stable since people’s stomach for risk changes. “After a bad shock, they redo their risk tolerance,” Smetters said. Part of the advisor’s job, he said, is to help people understand the markets.

Today, robos are a fast growing trend but they still command a small share of the market. In 2016, robos managed $126 billion in U.S. assets out of $69 trillion in assets under management globally, Fisch said. In addition, she said, a survey showed that 55% of people also don’t know anything about them. But that will change. In 2017, the Securities and Exchange Commission for the first time included robos as part of its “examination priorities,” according to a PwC report. Robos are regulated: They have to register as financial advisors, and as brokers-dealers if needed. They also have a fiduciary duty to seek a client’s best interest.

The 50-and-over Market

Robos are best for investors who are comfortable using a digital interface with minimal to no human contact. That means its clientele will skew young. “Most robo-advisors are focused on millennials and Gen Xers who are still accumulating assets,” said Elizabeth Kelly, senior vice president of operations at United Income. But it’s a “tough business model” because the cost to get a customer is high and the payback — if robos are paid based on a percentage of assets — is less because many younger clients haven’t accumulated that much yet.

It is the 50 and older crowd that holds 80% of investable assets, Kelly said. But too often, Silicon Valley is not innovating with the older group in mind. Most of the products designed for 50 and up are “gray, big and they’re ugly and not inspiring,” said Rhian Horgan, founder and CEO of Kindur, a retirement planning startup. One reason is “anyone over 50 is grouped together,” she said, when digital skills can vary widely between someone who is 50 and a senior at 90. “Baby Boomers know how to use computers and mobile phones but use them very differently than young folks.”

The opportunity in financial advisory services is to serve Baby Boomers, many of whom face complex retirement decisions as they figure out when to retire, how much money to withdraw and whether they’ll have enough money to last their whole lives. “There’s a huge amount of stress around this retirement decision,” Horgan said. A survey showed that 61% fear outliving their assets more than death, said Julianne Callaway, strategic research actuary at insurer RGAx. “There’s a tremendous cost to consumers for not planning for their end of life.”

It is the 50 and older crowd that holds 80% of investable assets. But too often, Silicon Valley is not innovating with the older group in mind.

Financial advice does pay off. Hammond cited research showing that investors who take professional investment help — including putting money in target-date funds — see their median annual return go up an extra 3-percentage points after fees compared to do-it-yourselfers. But most people don’t use financial advisors. And only 4.5% go for robo-advice, he said.

Designing with Nudges

How does one get people to change their investing habits? Behavioral science can help. For example, automatic enrollment in 401(k) plans with a choice to opt out makes people more likely to save, compared to having them opt in, said Jim Guszcza, U.S. chief data scientist at Deloitte. The secret is to work with people’s natural tendencies instead of forcing them to change. “Instead of an uphill climb, let’s make it a downhill climb,” he said.

“Instead of trying to make people more ‘rational,’ let’s just go with human psychology as a given and shape the choice environment around them. There are ways of changing behavior other than just giving people information they didn’t have before, or giving them economic incentives,” Guszcza added. “Subtle tweaks in the environment, the way choices are arranged, the way the information is communicated, can have surprisingly large impacts on people’s behavior.”

Years ago, Guszcza said, Deloitte worked with the state of New Mexico to monitor people taking unemployment benefits. The state wanted to find folks who are under-reporting income from side jobs and other work because this income would be deducted from their unemployment checks from the government. New Mexico wanted Deloitte to build machine learning algorithms to find problematic accounts so they could stop the benefits.

But it’s tough to design algorithms that would only catch cheaters. Guszcza said two out of three people flagged would be legitimate filers, and so cutting off their checks could have dire results. His team decided to use behavioral science instead to solve the problem. The algorithms would first look for anomalies — say, a benefits recipient would suddenly log on at a different time and report a much smaller income. That person would get a pop-up message that said, “Nine out of 10 people in your county accurately report their earnings every week.”

“It worked like gangbusters,” Guszcza said. Improper payments were cut in half. It turns out that people like to follow social norms; they want to belong and not be the odd one out. The team borrowed a page from Nobel laureate Richard Thaler and his work in behavioral science. Years ago, Thaler advised the U.K. government on how to get people to pay taxes on time. By adding a similar line on top of the letter sent to late-filers — nine out of 10 pay their taxes on time — the U.K. collected millions of pounds of additional revenue, Guszcza said.

“One of the [advantages] human advisors bring is to tell you, ‘That’s not what you should be worried about. You came to me about retirement and yet you have this huge credit card debt.’”–Kent Smetters

Shortcomings of Robos

While robos have their advantages, certain shortcomings can make them a deal breaker for some folks. One is what Fisch calls the “warm body effect.” “There’s some value in a complex financial world to having somebody to listen to you,” she said. “Somebody to talk you down when there’s a downturn in the market.” Empirical evidence shows that “access to humans can make people less risk-averse [and make them] less likely to overreact to market downturns — pulling their money out of the stock market at the bottom,” Fisch said.

Another shortcoming is that robos don’t educate clients about their investments. A human financial advisor can help customers understand their investments, leading to better decisions, Fisch said. The human advisor also can probe deeper than the robos’ questionnaires to assess a client’s changing needs and risk tolerance levels. “Human advisors can help customers come to more informed decisions about the level of risk they can, and should, bear,” she said. For example, investing too conservatively could mean inadequate funds at retirement.

Smetters added that human advisors are also more able to give clients the big picture. “One of the [advantages] human advisors bring is to tell you, ‘That’s not what you should be worried about. You came to me about retirement and yet you have this huge credit card debt,’” he said. Also, robos focus on asset allocation when other factors — such as saving habits — have a greater impact on whether one will have enough for retirement, he noted.

Robos also are better for investors who are in the process of accumulating wealth than those in the “decumulation” stage when they are withdrawing money from retirement accounts, said Steve Polansky, senior director of special initiatives and shared services at the Financial Industry Regulatory Authority (FINRA). Retirees face a host of issues such as which accounts to withdraw from first to maximize tax benefits, life expectancy, medical costs, inflation, long-term care and others. These are questions robos are not programmed to handle adequately.

The fundamental drawback of robos is that they’re still just software — and “computers are only as good as the information that you put into them,” Fisch said. “Even if a computer program is excellent, it doesn’t necessarily identify and draw out all of the information that’s relevant to somebody’s financial situation. A human advisor is more flexible and often better able to do that.”