Wealth managers face a new challenger: their clients’ AI chatbots

A version of this article first appeared in CNBC’s Inside Wealth newsletter with Robert Frank, a weekly guide to the high-net-worth investor and consumer. Sign up to receive future editions, straight to your inbox.

Even high-net-worth clients who can afford top-notch advisors are asking artificial intelligence chatbots like Claude for portfolio recommendations and tax advice, wealth management leaders told CNBC. 

“My personal opinion is that ChatGPT is the single largest investment advisor in the world right now,” said Matthew Fleissig, CEO and cofounder of Pathstone, a registered investor advisory with $185 billion in assets.

Asking AI for a second opinion can help clients come up with informed questions and have deeper conversations with their financial advisors, according to firm leaders. That said, the practice comes with risks, such as getting incorrect advice or having personal information leaked.

“I think for a client who’s dealing with something that’s very technically complex and doesn’t have that grasp, it may be harder to differentiate between a hallucination or an error of fact versus a good insight that the engine has,” said Michael Zeuner, managing partner at WE Family Offices.

Moreover, as these large language models get more sophisticated and popular, wealth advisories – especially those that target the mass affluent — will have to do more to justify their fees, according to Morningstar’s Sean Dunlop. He told CNBC that wealth management stocks have already pulled back as AI has encroached on the industry, such as an AI tax planning tool by Altruist released in February or a personal finance feature launched by OpenAI in May.

Dunlop said it’s unlikely that AI will wipe out traditional wealth managers — but they will do more than help advisors become more efficient.

“I think the truth’s probably somewhere in the middle where the service level is going to get better. You probably need fewer advisors to serve the pool of assets, which might itself expand a little bit, and there’s going to be some group of customers that are willing to do it themselves that weren’t before,” said Dunlop, director of equity research at Morningstar. 

“At a minimum, it ought to raise the floor. Like, if you’re an advisor and you’re keeping half your client’s balance in cash in an IRA, then this really ought to be a wake up call,” he added.

Pamela Lucina of Northern Trust said she first noticed clients using AI to double-check the firm’s advice about 18 months ago and that it has become a more frequent occurrence since. 

“We’ve had clients tell us directly that they’re going to ask AI the questions that they’re going to ask us,” said Lucina, the firm’s chief fiduciary officer and leader of its trust and advisory practice. “I think what they’re often testing us for is not the answer, but having more specifics or evidence that we’ve actually done these things before.”

Many prospective clients also use LLMs to help them decide whether to work with Northern Trust, she added. Before the rise of AI chatbots, typically only billionaire clients would ask the firm to submit a formal proposal for managing their wealth, according to Lucina. Now she estimates about half of clients send requests for proposals, even those with as little as $100 million in assets. Some have told the firm that they used ChatGPT or other LLMs to formulate their highly specific questions, she said. 

This process can save time and make client meetings more efficient, Lucina said.

“Less of our conversation is about information sharing and more about focusing on the outcome they are trying to achieve and how it actually applies to families,” she said. “The fact that there’s more information out there is not a threat. It can take a conversation to the next level.”

Zeuner said he first noticed clients talking openly about feeding portfolio recommendations into LLMs about five months ago. It has become so common that his firm is considering whether to run every recommendation in Microsoft Copilot so advisors can prepare for questions clients may ask.

He said he views the use of LLMs as a second set of eyes and a positive trend.

“The AI is enabling and empowering them to get involved, to learn, to ask questions, to get a second opinion,” he said. “Whether they challenge us or validate, it doesn’t matter; they’re engaged.”

Still, Zeuner said he’s concerned about clients not catching AI-made mistakes. For instance, he said, if a client uploads a trust document into a LLM and asks a question about it days later, the chatbot might hallucinate details of the document in its answer. 

Zeuner also said he’s seen LLMs make simple mistakes, such as ChatGPT telling a client that two ETFs in his portfolio appeared to be identical. While the ETFs tracked the same index, one was equal weight and the other was cap-weighted, he said.

When Lucina asked an LLM for examples of how clients could save on capital gains taxes, it got the math completely wrong, she said.

“I generally use AI to poke holes, to try to find criticisms in a way of thinking, and I think it is good for that,” she said. “But more often than not, especially on the advice side, it’s wrong.”

 More crucially, clients risk exposing identifying information if they use a personal account, even if they spring for a paid plan. Financial institutions, including wealth advisories, typically pay for enterprise plans that come with strict data protections. For instance, WE Family Offices has an agreement with Microsoft Copilot that data the firm enters is never used to train public AI models, according to Zeuner.

“We talk to [clients] about general best practices for using these LLMs, given their high profiles and wealth,” he said. “If you have an AI assistant capture every meeting, where is that transcript going? Do you know how it is stored? Who has access to it and how secure is it?”

Fleissig said that while AI chatbots like Anthropic’s Claude and OpenAI’s ChatGPT are capable of great analysis, that’s not the same as actionable advice. 

He described LLMs as “semi-objective tools” that aren’t always equipped to make subjective decisions. A trusted advisor can guide clients to make the choice that’s right for them based on a variety of factors, he said.

“There is value in human-to-human interaction from a relationship standpoint because there’s so much nuance to it,” he said.

And though an LLM could make an argument for investing in pre-initial public offering Anthropic, for example, it can’t put that advice into practice and get you access to a funding round.

“You get access because of relationships that you have, and you also have all these experiences that you’ve had with clients,” Fleissig said. “Deals today are still not getting done via AI.”

Vince Lumia, a Morgan Stanley veteran of 27 years, said that the value of human touch is more easily overlooked during bull markets. That changes during times of crisis or uncertainty, said Lumia, who oversees Morgan Stanley’s 16,000-plus financial advisors. 

“The world sort of starts to think that it’s actually easy. The one guarantee I could have is that the markets will be volatile, and they won’t go up in a straight line, and there’ll be disturbances,” he said. “And when those disturbances happen, larger clients tend to want advice.”