AI wants to spend your money for you. Should you trust it?

MW AI wants to spend your money for you. Should you trust it?

By Genna Contino

From virtual shopping assistants to filing taxes, there are countless ways to integrate AI into your personal finances today

Google, Walmart, Mastercard and other companies want you to use agentic AI for financial transactions. Here's when this makes sense and when to steer clear.

Financial planner Stephen Kates thinks using artificial-intelligence tools for personal finances is similar to using a chainsaw: "In the hands of the right person with the right capabilities, you can carve a beautiful statue or cut a tree or do something very useful."

For someone who doesn't know what they're doing, however? "They can lose a limb," said Kates, a financial analyst at Bankrate. "And you could do the same to yourself financially."

If it feels like AI tools for managing your finances are everywhere lately, you're not imagining it.

Opening your banking app to check your balance might result in an AI assistant promoting a high-yield CD you never asked about. TV commercial breaks and Instagram algorithms are peppered with ads for AI-fueled platforms experimenting with AI that promise to cancel your unused subscriptions or find missed deductions in your taxes. And people are trying out these new tools: 51% of consumers have turned to AI for financial advice, and 27% said they've considered it, according to an August 2025 report from JD Power.

But the latest AI technology expands beyond chatbots spewing out financial advice - there have been recent developments in agentic AI, which is when a system acts autonomously on a user's behalf.

As AI's capabilities shift from chatbot to autonomous executor, the stakes shift from tech glitches to major life disruptions. It's one thing for a bot to suggest you save more and another for a bot to have access to your checking account and the power to spend from it. Some experts believe these agents aren't prepared to take on these risks. Here's what consumers should know about these tools, and one key rule to follow to protect your money and your identity.

"Agents aren't ready for prime time in doing the kind of execution on financial tasks that carry a lot of weight or importance," Kates said. "We still have mistakes. And that's something that people should not tolerate. They should not tolerate an agent with the ability to make a mistake on something as important as their money and their future."

Read more: How AI could drive a renaissance for blue-collar workers

A recent example of agentic AI is rewards program Bilt's new AI "neighborhood concierge," which books restaurants and schedules rent payments for Bilt users. Walmart $(WMT)$ recently partnered with both OpenAI and Alphabet's $(GOOGL)$ $(GOOG)$ Google on a tool that lets shoppers plan meals, restock essentials and complete purchases simply by chatting with a bot. Mastercard $(MA)$, meanwhile, built a payment system called Agent Pay, designed to let AI assistants make purchases on users' behalf at millions of retailers.

While shopping bots are already clicking "purchase," traditional financial powerhouses are moving more cautiously. Experian (UK:EXPN) this week launched EVA, an AI assistant designed to emulate a human credit expert. The goal is to help users understand and improve their credit scores. But for now, the tool stays on the side of information rather than action.

"EVA currently provides guidance on what actions the user could take, and doesn't take action on the user's behalf," says Debbie Hsu, executive vice president of product at Experian Consumer Services. While EVA can flag large transactions or explain why a credit score changed, it won't actually freeze your credit or pay a bill for you - at least not yet. Hsu said Experian hopes to "build more agentic capabilities" for the bot in the future.

The agentic-AI trend has put federal regulators on high alert. In August 2025, the Federal Trade Commission filed a landmark complaint against a company called Air AI, accusing it of over-promising what its autonomous bots could actually do. The agency's concern was that while these tools are marketed as flawless replacements for human experts, they are often prone to making costly, unauthorized purchases.

The stakes are lower with something like an agentic concierge - you might have to pay a fee if you miss a scheduled restaurant reservation, for example. But Kates is worried about a world where AI agents become mainstream among companies that help with more consequential financial services, such as tax preparation or investing platforms.

Don't Short Yourself: ??The smart way to use AI to do your taxes

Relying on large language models that can pull from outdated tax laws to file your taxes, for instance, when "you don't have the ability to double check or fact check these things," Kates said, "you can get into trouble."

Though investing platforms like Robinhood (HOOD) and Fidelity have integrated AI throughout their apps, and there are many AI-assisted tax-prep services, no widely adopted app for these services powered by agentic AI exists yet. But that hasn't stopped some AI enthusiasts from attempting to build their own workarounds by training large language models like ChatGPT, Claude or Gemini to prepare tax returns or trade stocks on their behalf.

Where AI tools genuinely shine, Kates said, is as a "second brain," or a sounding board for financial strategy, rather than an executor of it. People can ask chatbots to model out goals, stress-test a savings plan or explain a concept they don't fully understand. "Have it pick holes in the strategy you currently have," he said. "Get into that back and forth."

But it's important to recognize that AI can only respond based on what it's given, and it doesn't understand personal priorities or emotional nuance.

From the archives (July 2024): I asked AI and my financial planner the same questions. Here's how they stacked up.

"That's where human judgment still matters most," said Thomas Racca, manager of the Navy Federal Credit Union personal-finance management team. "Decisions around investments, retirement planning, debt strategy or major life events, like buying a home, require personalized guidance, experience and an understanding of someone's broader financial picture."

When interacting with any large language model, Racca said, make sure you don't upload Social Security numbers, passwords, tax documents or any other information that could expose your identity or give access to your financial accounts.

What personal-finance issues would you like to see covered in MarketWatch? We would like to hear from readers about their financial decisions and money-related questions. You can write to us at readerstories@marketwatch.com. A reporter may be in touch to learn more. MarketWatch will not attribute your answers to you by name without your permission.

-Genna Contino

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

March 19, 2026 15:55 ET (19:55 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.