Recently, as an experiment, I gave a bunch of my financial information to an AI chatbot. I know what you’re thinking — it sure doesn’t sound like the wisest move. While the large language models that power such bots are good at many things, math is not historically one of them. Still, a growing number of startups are training generative AI tools to work like financial coaches. Some of them are pretty good, too.
Cleo, which bills itself as “the world’s first AI financial assistant,” got a big update that uses OpenAI’s sophisticated o3 reasoning model to help users break down complex financial decisions. The app also now has the ability to remember your financial history and goals. You can even talk to the app with a new two-way voice feature. At some point, Cleo might ask you if you want to take out a loan.
I’d been using the previous version of Cleo for a few weeks without any big breakthroughs. But when I linked my bank accounts (securely!) to Cleo 3.0, the latest version of the artificially intelligent financial assistant, I was stunned by how handy it was to ask detailed questions about my spending habits, savings goals, and retirement planning in plain English and get useful answers. The app is built on top of ChatGPT and is specifically trained to handle those kinds of questions. It’s also equipped with tools to do the math correctly.
In the past, only the very wealthy have been able to map out their financial futures like this. In the very near future, anyone could do it for free.
There’s something revolutionary about this concept. And that’s not just Cleo’s executives talking. A number of experts have told me that tools built on top of large language models like ChatGPT stand to transform the world of financial advice and planning. They have the ability to ingest vast amounts of your financial data, they’re trained on the same materials a human financial planner might use, and they can answer an endless number of questions based on that knowledge and expertise. In the past, only the very wealthy have been able to map out their financial futures like this. In the very near future, anyone could do it for free.
“We’re on the cusp of a pretty significant change in how [people] are able to use AI to help manage their finances,” said Andrew Lo, a professor of finance at the MIT Sloan School of Management. “I don’t think these apps are all ready for primetime, but nonetheless, there is a sea change between today’s AI finance applications and what existed even two years ago.”
Such a sea change comes with its own challenges. The arithmetic issue is one of them, although there are ways to build these apps in order to curb hallucinations. Another is ensuring the right guidance is tailored to the right user — what’s called “suitability” in the finance world. But, Lo explained, “probably the most important and most difficult challenge is trust and ethics.” How do you ensure an AI fulfills its fiduciary duty to give you the best advice? Licensed financial advisers, who work in a highly regulated industry, can face civil or even criminal charges for failing their clients. An AI currently cannot.
That means you definitely should not let an app talk you into pouring your retirement savings into meme stocks. But you might consider letting a chatbot take a peek at what you’re spending on streaming services. Did you know, for instance, that canceling Apple TV+ and putting that $10 a month into an IRA could add up to over $12,000 saved in 30 years? That’s a fact I learned from Cleo.
The artificial lightness of banking
My first brush with algorithmic financial advice was a decade ago, when I downloaded an app called Digit. The proposition was simple: You connect your bank account to Digit, which would analyze your spending and strategically slide a few dollars or cents into a savings account. I ended up saving thousands of dollars using Digit without really noticing, which was precisely the point.
There’s a lightness to the Digit experience. Contrast that with Mint, the now-defunct app and website that let you connect to your bank accounts and used machine learning to categorize your spending. Mint was heavy, because it required a lot of time to set up; you had to make sure all the automation was working correctly and then had to keep the settings updated as your financial situation changed. An array of finance and budgeting apps have swept in to replace Mint after its 2024 closure: YNAB, Monarch Money, Origin, to name a few. They’re heavy, too, although they’re increasingly incorporating AI to make things run more smoothly.
Most of these apps use a service called Plaid in order to keep your financial data secure. Plaid, which is used by companies like Venmo, Robinhood, and Chime, gives the apps real-time, read-only access to your accounts so the apps never have direct access to your money. There’s always the risk of a data breach, which is not unheard of in the fintech world.
While Plaid has made it easier for apps to see all your balances and transactions, what people actually need from a fintech app varies widely.
“If the product can help you make a complex financial decision confidently and competently in minutes, instead of hours with a bunch of pain and stress and mental suffering, more people will use it.”
— Ethan Bloch, Digit founder
“Personal finance is incredibly niche, because each person, on a set of dimensions, is a unique snowflake financially,” Ethan Bloch, the founder of Digit, told me. Bloch sold Digit to the financial technology, or fintech, company Oportun in 2021 and is currently working on a new AI-powered financial tool called Hiro, which promises to “turn your financial data into personalized advice” through a chatbot. Bloch said that large language models have made this possible and that, like Digit, the experience can be incredibly light and feel effortless.
“If the product can help you make a complex financial decision confidently and competently in minutes, instead of hours with a bunch of pain and stress and mental suffering, more people will use it,” he said.
Of course, this is assuming the AI never gets anything wrong.
Fintech apps are not banks — which means that even though they might handle your money or your financial data, an app like Cleo is not as heavily regulated as a chartered bank, like Chase or Bank of America. Fintech apps might require licenses to take part in certain activities, like moving money between accounts or holding balances. If an app gives investment advice, it might face scrutiny from the Securities and Exchange Commission (SEC) or state regulators. For the most part, though, there is no regulatory oversight when it comes to financial coaching, which is what Cleo does.
Making sure AI-powered fintech apps don’t dupe Americans was a job for the Consumer Financial Protection Bureau (CFPB), but it’s unclear how much power the agency has under the Trump administration. For example, the CFPB passed a rule last year that established a set of personal financial data rights around what happens when you connect your bank account to an app like Cleo, YNAB, or Monarch Money. That rule is now being rewritten under the Trump administration, as congressional Republicans attempt to defund the CFPB altogether.
“The vision that came out of the financial crisis was that we would have a strong regulator putting consumers first, because for too long, the regulators had put banks first,” said Aaron Klein, a senior fellow at Brookings, “and that vision has been destroyed by the Trump administration.”
It almost goes without saying these days that generative AI is a new technology, and the apps using large language models for financial advice are even newer. There is not a lot of oversight of this space — there’s not none — but you should be careful when entrusting your financial fate to a chatbot.
The proposition is still intriguing. We’ve already seen AI chatbots streamline the tax-filing process, and it’s increasingly clear that AI can make you better at your job, as long as it’s not stealing it. It seems very possible that in the near future, AI will make it easier to make a budget and stick to it. It might help you plan better for your retirement. I am not yet asking an AI chatbot for investment advice — there is just too much downside when it comes to sensitive decisions involving large sums of money. But talking through ideas for how I might get more out of my money and then fact-checking everything the bot tells me? That makes sense. At the end of the day, I still want to talk to living, breathing experts about my big financial decisions.
But for the smaller stuff, this technology could help a lot of people. It could democratize financial advice, in a sense.
“On the bright side, I think that a large number of individuals, who are currently not getting any financial advice and badly need it — they will have access to pretty good financial advice at no cost,” said Lo, the MIT professor. “That’s the promise of AI over the course of the next few months, not to mention years.”
A version of this story was also published in the User Friendly newsletter. Sign up here so you don’t miss the next one!
