Close Close

Financial Planning > Behavioral Finance

As Advisors Jump Into AI, Experts Warn About Data

Your article was successfully shared with the contacts you provided.

Artificial intelligence (AI) is seen as becoming an increasingly important tool for financial advisors, but thought leaders in the field urge firms to use caution despite the lure of potential rewards.

It is likely that AI – as it continues to be developed – can help financial advisors better understand a client’s risk profile and spending habits.

“Advice can be provided faster [with AI],” says Reena Aggarwal, a professor at Georgetown University, where she directs the Center for Financial Markets and Policy.

“AI [also] can help clients make better decisions.”

Yet, there are concerns. On one level, investment advisors need to be careful about the quality of the data and algorithms being put into use.

(Related: What Role Could AI Play in Investor Relationships?)

“AI can only work when you have good data,” Aggarwal warned. So, the data must be timely, accurate and correct, she said. Some AI can be predictive and that means using assumptions – which could be inaccurate, she adds.

“In terms of the actual advice that financial advisors provide, the use of AI could certainly help, but the advisors should be more cautious and stay aware of the limitations of AI,” Jillian Grennan, a professor at Duke University, agreed.

Based on her recent research, Grennan explained that data scientists are making the “implicit assumption” that the data-generated process itself is not changing over time. “But … the underlying data-generating process is being affected by these fintechs,” she suggests. She found that “sell-side equity analysts produce more biased, less accurate reports where fintechs concentrate.”

“I think the challenge for financial advisors will be to figure out if the AI they are using may in fact change the data-generating process,” she said.

So, firms need to know if fintechs that provide signals, or use AI to provide signals, are updating their algorithm continuously, she said.

Vasant Dhar, a professor at New York University, even questions if advanced offerings are yet available – though he says they will become more prevalent – where products are tailored through AI and are matched with investors.

“The systems really don’t exist,” Dhar said. “If it does, I would be highly skeptical.”

Dhar, who developed the Adaptive Quant Trading (AQT) program for SCT Capital Management, a hedge fund based on machine learning, said that his offering is different from a type of offering   where products are individualized to a client’s risk profile.

Dhar further urged investment advisors to be skeptical about adopting many AI offerings – and especially look at the scientific background, credibility and motivation of the developers. If social media is used, he warns there could be manipulation risks.

“Really kick the tires and do due diligence,” Dhar advised. “It’s an exciting area, but an area where people should be really cautious…. It’s important not to get caught up in hype.”

There are also concerns about protecting and monitoring the large caches of data used in AI, which can get corrupted and be impacted by biases. There are risks that hackers or other nefarious actors can alter the data or algorithms – or someone can even make unintentional mistakes, according to Kevin Petrasic, an attorney at White & Case.

While financial advisors do not need to understand the precise details used by the data scientists in coming up with the AI algorithms, they need to have enough of a working knowledge so they can be transparent with their clients and have an appropriate level of integrity, Petrasic said. It is part of the financial advisor’s job to explain the risk to clients, he said.

Financial advisors should understand what an algorithm is designed to do, he added. That means “understanding the pros and cons of analytical solutions they [the advisors] are relying upon,” he said.

Moreover, firms do not want to have algorithms colluding with each other based on timing and manipulation, according to Petrasic. There could be concerns, for instance, if one program becomes the leading program and others follow that program. Moreover, there could be issues when an off-the-shelf product is improperly connected with a proprietary product.

In fact, if a firm fails to undertake a sufficient level of due diligence, it can trigger a government investigation by an agency like the Financial Industry Regulatory Authority, Petrasic adds. Or, it could lead to a legal action by investors who lose money.

But despite the risks, Aggarwal said AI “is here to stay.”

AI can help with customer engagement, make predictions about customer needs, and identify potential clients from alternative data such as social media, according to Grennan. For example, Gridspace uses AI to analyze voice patterns in real-time and determine the customer’s mood during service calls so automated scripts can be adjusted, she said. “A similar-type of AI could be very useful for financial advisors with their clients,” Grennan adds.

She says AI will be most successful, initially, when determining risk. “That is incorporating social media and other information to identify a clients’ risk-preferences,” she explained. Alternative data can also be employed to detect fraud, Grennan said. 

Overall, given its many benefits, “most firms are adopting [AI] in some form or another,” Aggarwal said. “Firms just have to figure out how best to use it… Everyone wants to have a competitive advantage.”

(Related: What Role Could AI Play in Investor Relationships?)