AI could soon act as ethical financial advisors, MIT researchers predict

• MIT researchers say it’s possible to train AI to be a trusted financial advisor.
• But it should be able to understand emotions to build trust with users.
• This article is part of “Build IT,” a series on digital technology trends that are disrupting industries.

Arynton Hardy, a wealth manager in Los Angeles, uses artificial intelligence every day.

“Almost every meeting I have with a client, I’m using an AI summary that gives me notes, gives me follow-ups, gives me tasks to do,” Hardy said. His employer, Savvy Wealth, is a technology startup focused on delivering AI innovations to its growing list of financial advisors and high net worth clients.

Hardy said AI tools regularly save his team hours on data entry, portfolio monitoring and other back-office tasks, giving him more time to meet with clients.

He’s not alone: ​​A report that data analytics firm Escalent shared with Business Insider said that nearly 40 percent of financial advisors have used generative AI tools in the workplace, primarily to increase productivity, a generate content and market or to prospect new customers.

Soon, generative AI may have the power to fulfill the most important role of a financial advisor: giving people reliable money advice.

While adoption of ChatGPT and other large language models has been rapid in industries including fashion and marketing, financial services have hit regulatory roadblocks.

But MIT researchers believe there is a clear path to training AI models as subject matter experts, ethically tailoring financial advice to a person’s circumstances. Instead of answering “How should I invest?” with generic advice and an urge to seek professional help, an AI chatbot could become the financial advisor itself.

“We’re on our way to that Holy Grail,” said Andrew Lo, professor of finance at the MIT Sloan School of Management and director of the Financial Engineering Lab. “We think we have about two or three years before we can demonstrate software that, according to SEC regulatory guidelines, will satisfy the fiduciary duty.”

Fiduciary duty refers to the set of legal responsibilities that implore financial and investment advisors to exercise the highest degree of care when handling a client’s money. It is the gold standard in financial planning and has so far proven difficult to replicate in direct-to-consumer AI tools.

AI is ‘inherently sociopathic’

Financial advisors often develop recommendations for clients through a behavioral finance lens, as research suggests that people do not always make rational or unbiased financial decisions, but are error-prone and emotionally motivated.

An average investor, for example, might panic when the stock market crashes and decide to sell their holdings in a mutual fund rather than wait for the market to recover, as it almost always does.

Knee jerk reactions like this can often be avoided or corrected under the guidance of a qualified financial advisor that the investor trusts. Lo said trust can be developed in part through empathy and small talk.

“When you start talking to someone, you almost immediately develop feelings for that person,” Lo said. “This is the kind of process that must happen with large language patterns. We need to develop an ability to interact with people not just intellectually, but emotionally as well.”

But the glaring problem with publicly available AI tools is that they are “inherently sociopathic,” Lo and his co-author wrote in a research report exploring the challenges of widespread adoption of AI-based financial advice.

“This sociopathy appears to cause the characteristic weakening of LLM results; an LLM can easily argue both sides of an argument because neither side carries weight,” they wrote. It may be able to play the role of a financial advisor based on its training data, but AI needs to have a deeper understanding of a client’s mood to build trust.

One way to achieve this is by having the machine ask the user simple questions, starting with “How are you?” before giving personalized financial advice.

In addition to text analysis, advanced AI could be able to request audio or video from the user to identify emotional cues, such as stress or fear, in the voice or facial expressions, Lo said. Think of it as a doctor displaying the kind of bedside manner that is praised in the medical field.

“Trust is not something that will be automatically given to a generative AI,” Lo told BI. “It must be earned.”

Planning for optimization

Conquest Planning, a Canadian startup, has developed a financial planning software that uses an AI architecture known as the blackboard system.

Ken Lotocki, its chief product officer, said the “blackboard” stores essential information about tax rules, cash flow mechanisms, retirement account structures, fiduciary rules and more.

Customers enter their personal and financial data so the system can learn their preferences, parameters and goals. Using that information, Lotocki said, the system “looks at all the permutations and options available” and then recommends a course of action.

Advisors can switch between a list of strategies to measure the effects of different decisions on the client’s wealth and goals, such as claiming retirement benefits at different ages.

Mark McGrath, a financial planner and portfolio manager at PWL Capital in Canada, said the software saved him time and provided clients with better financial plans than he could create by hand.

“I’ve been doing this for a long time — I assure you, no financial planner goes into that level of depth and spends that amount of time optimizing,” McGrath said. “That means customers don’t always have an optimal financial plan.”

He added: “The analogy I’ve used in the past is that with the Conquest I’m driving a professionally tuned Ferrari while my competition is on a pair of roller skates.”

Lotocki estimated that Conquest’s software is used by 50 percent to 60 percent of the financial advisor market in Canada through partnerships with banks and independent consulting firms. The software is available in the US on Pershing X’s wealth management platform, Wove.

Although sophisticated, the software is not a direct-to-consumer solution. Lotocki and his associates hope to help make financial advice AI accessible to people across the wealth spectrum, perhaps for free.

Financial planning 3.0

Many financial advisors are eager to use generative AI as an assistant, but few are ready to replace them.

Lo said he believes a world where people rely on AI advisors rather than human advisors is on the horizon. But he said a smooth transition would require advisers to retrain for new careers, possibly with government support.

“What worries me, and what I think policymakers really need to focus on, is if a large body of human employees become displaced in a very short period of time. This could cause tremendous social unrest and dislocation,” Lo said. “So the speed of movement is something we have to be careful about.”