Paul Weiner, an artist, has been experimenting with artificial intelligence for the past year, generating A.I.-created visual disinformation and seeing whether he can get the images to spread. But recently, he turned to ChatGPT, a chatbot that has the ability to respond to complex questions, for a much different reason: With his 30th birthday looming, he decided to ask it for advice about retirement planning.
“Maybe ChatGPT would have some answers that I might otherwise get from someone who I’d have to pay a lot of money to,” he said.
Generative A.I. like ChatGPT has knowledge workers gripping the rails, bracing for how it might affect their jobs, and consumers leaning in to see what costly services could soon be replaced with a prompt. As the investment industry turns to artificial intelligence as a financial planning and advice tool, the values of accuracy, humanity, security and accessibility are jostling for prominence. In the future, who — or what — will we be asking to advise us on some of life’s most important decisions?
ChatGPT recommended that Mr. Weiner open a Roth individual retirement account and certificates of deposit, as well as automate his savings and create a budget. He hasn’t yet opened any of the accounts or, as the chatbot also suggested, worked with a financial adviser.
“It’s a lot of information that gets thrown at you pretty quickly,” Mr. Weiner said. He found the short explanations insufficient for what a C.D. does or the differences between a Roth I.R.A. and a traditional I.R.A. He concluded that speaking to a financial adviser would probably be more helpful.
“But that kind of circles back to the whole reason I’m doing this on ChatGPT to start with — it’s free,” he said.
A.I. has joined the financial chat
Delyanne Barros, a money coach, said she felt that most of the hundreds of thousands of people who follow her on social media had no idea what ChatGPT is. “Am I the only one geeking out on this thing?” she asked. When she asks her followers if they’ve used it, she said, “they’re like, ‘What are you talking about?’”
She’s teaching them the basics: There’s a free version of the service, and it works as more than just a Google alternative.
On Instagram, she asked if any investing newbies had asked ChatGPT to teach them to invest. Some had tried but reported that they kept getting stuck in a loop of repetitive answers. Ms. Barros found that she was able to get valuable information about allocations, tax efficiencies and retirement withdrawal rates, but she posits that was because she had knowledge of the investment terms she needed to use.
“You have to know how to frame the questions,” she said. “A lot of people don’t understand that you get an answer to something and it can build on that answer. You can ask follow-up questions, and it’s like a chain.”
Ms. Barros has also used ChatGPT to double-check her calculations regarding her retirement plan. Despite its handiness, she is not worried that chatbots will replace her.
“With something like investing, I’m not concerned as a personal finance educator, because I can see that it’s not like: ‘Oh, we don’t need you anymore. We have ChatGPT,’” she said. “If anything, this is going to be a tool that’s going to enhance my coaching experience with people, but it’s definitely not going to be replacing us, because people still need a lot of guidance.”
Even if you don’t think you’re familiar with it, chances are you’ve already been using generative A.I.
Intuit started to integrate A.I. into its software products, which include Mint and TurboTax, more than a decade ago, said Ashok Srivastava, the company’s senior vice president and chief data officer. Today, he said, Intuit’s platform performs 58 billion machine learning predictions per day. Another Intuit product, QuickBooks, predicts cash flow for small businesses, and the company has found that when it gives users advice based on artificial intelligence, 95 percent of small-business owners take that advice.
They’re still focusing on a strategy that combines human interactions with A.I.-powered ones. Customers, for example, can meet with a live expert, and then A.I. will create a categorized and tagged summary of the conversation for later review.
Bugs in the system
As of now, the technology is promising, but it’s not 100 percent accurate.
“These systems tell plausible stories, they give you plausible ideas, but not necessarily correct ones,” Mr. Srivastava said. “What we’re focusing on is actually providing the correct experience to the person, so that it’s grounded in reality and data that is appropriately personalized to them, so then they can make the best financial decisions as they move forward.”
Mr. Srivastava said he did not envision a future where humans were taken out of the financial planning equation.
“I’ve grown up in the field, I’ve seen it evolve, and it’s an amazing technology,” he said. “I think that the human connection is still important. I envision that we will want to help C.P.A.s, bookkeepers, financial planners, financial advisers — everyone in this ecosystem — grow and prosper along with the use of artificial intelligence.”
Josh Pigford, the founder and chief executive of Maybe, had been building a personal finance management platform that could help people make financial decisions when ChatGPT debuted. A few months ago, Maybe was rebuilt from the ground up, this time with GPT, the technology behind ChatGPT, as the foundation of the platform. The process always begins, he said, with a question people want to answer.
“The way that we were initially tackling this is giving you access to a financial adviser who can answer those questions for you directly,” Mr. Pigford said. “As we started testing GPT’s ability around that, we realized, well, OK, actually GPT can do this really well.”
Things became even more interesting when people added their financial data and information, such as age, location, and goals. The system could then take into account everything from dependents to joint filing to local tax codes — details a financial adviser would be able to use — and deliver that directly to the consumer.
That, of course, brings up the subject of privacy. Through Maybe’s system, the banking information is secured and does not feed back to OpenAI, the company that created ChatGPT.
Hallucinations — the tendency for ChatGPT to spout off incorrect information — have also become a worry. Mr. Pigford and his team identified the issue during early testing.
“There was a point there where it was actually making up entire transactions, and building this back story of like, ‘You bought this item from Home Depot to help cool off your living room,’” he said. “That’s a legitimate problem.”
As the technology has improved, Mr. Pigford has seen a drastic decrease in these hallucinations in just weeks. The way they’re designing the software includes a toggle to switch between a chatbot and humans for advice.
“The belief, the hypothesis, what we’re sort of banking on is that we’re able to actually offer that sort of hyper-personalized input and advice without you having to, you know, form a relationship with a certified financial adviser where you’re paying them an assets-under-management fee, or even paying them, you know, a couple hundred bucks an hour,” he said. “You’re able to get very specific advice, regardless of what your financial situation is.”
But Mr. Pigford believes it’s too early to do away with live professionals. “I think we’ll have some transition period where we’ll want humans involved for a while,” he said. “The goal is not to completely do away with a financial adviser.”
First steps into the ChatGPT world
Glenn Hopper, author of “Deep Finance: Corporate Finance in the Information Age,” relates this GPT era to the screech of dial-up internet. The prevalence of A.I., he said, is “going to come quicker than the adoption of the internet and broadband internet and web browsers.”
“I’ve stopped making predictions, because every time I make a prediction, I’ll say six to 12 months, and then I’ll read an article the next day that this item has already appeared,” Mr. Hopper said.
He warned that tools like ChatGPT would make scamming and phishing more sophisticated, so users should be cautious of anyone asking for their bank information.
“The very first thing that I tell everyone is, if you’ve been ignoring artificial intelligence up until now — stop,” he said. He doesn’t think people need to become experts, but they should have a basic understanding of how the technology works, he said.
“If we’re going to hand over our decisions to them, and we don’t have any idea how they’re working, I mean, you might as well shake one of those Magic 8 Balls and get the answer from that,” he said.