Two things are true at the same time this week. AI chatbots get specific money questions wrong about half the time, and Americans trust them anyway.
That would be concerning enough on its own. Then on Monday, ChatGPT started holding people's actual credit reports.
Your clients' favorite second opinion now knows their real numbers. Meanwhile, the $35 million Envestnet promised three weeks ago started showing up on advisor screens.
Three stories from this week that matter for your practice.
New Research Says AI Flubs Money Questions Half the Time. People Trust It Anyway.
On Thursday, Digital Wealth News pulled together the research wave that ran through the financial press this week, and the two halves don't seem like they should fit.
New studies from Saturn and DeepVest found that AI chatbots failed to correctly answer specific financial questions, the kind built on real-world scenarios, at least half the time. At the same time, academic research keeps finding the opposite result one level up. An MIT study from earlier this year showed AI guidance moving people toward the money habits most planners would recognize as good ones, and a Stanford study sponsored by Jump found something similar.
Both can be true at once. A chatbot can correctly tell someone to build an emergency fund and diversify, then botch the tax math or miss an account rule in the same conversation.
The trust numbers are the part worth reading twice. A New York Life survey from mid-September found 82% of Americans who sought financial guidance from AI rated the advice as good or better.
A PensionBee survey found 57% of American AI users would accept a chatbot's answer on a financial decision without checking it, even though about a quarter admit they've gotten wrong information from one before.
Why you should care: Your clients live inside this paradox. The tool in their pocket is genuinely helpful with concepts and habits, confidently wrong on specifics about half the time, and trusted by more than half of its users without a second look.
Telling clients to stay away from AI won't work, and the research says it would be bad advice anyway, since people who use it tend to make better money decisions overall.
The move that works is claiming the job the research just described. Someone needs to be the person who checks the specifics, the tax math, the account rules and the deadlines, before a client acts. That's you, and your clients need to hear you say it out loud.
ChatGPT Can Now Hold Your Client's Credit Report and Bank Accounts.
On Monday, OpenAI added credit scores to Finances in ChatGPT. Plus and Pro subscribers in the US can now connect their Experian credit report and see their VantageScore 3.0 right inside ChatGPT, next to the bank, credit card, loan and investment accounts they can already connect through Plaid.
Once connected, a client can ask ChatGPT why their score moved, which card is eating their available credit, and what to pay off first. The report refreshes monthly, sends alerts when something changes, and checking it counts as a soft inquiry, so looking doesn't hurt the score.
OpenAI even tells users to type their verification details into Experian's form instead of the chat window, which is sound advice from a chatbot about chatbots.
There are real limits worth knowing. The score shown is a VantageScore 3.0, which some lenders use and many don't, so the number in ChatGPT may differ from the one a mortgage lender pulls.
And if a client disconnects, OpenAI deletes the report data within 30 days, but anything already sitting in past conversations stays until they delete it themselves.
Why you should care: Put this next to the first story and the picture gets sharp. The tool that gets specifics wrong about half the time now has your clients' real balances, real payment history and real credit file.
Expect clients to show up with observations ChatGPT made about their own numbers, and expect some of those observations to be good. This is the same pattern as the research above. AI is helpful with the broad strokes and shaky on the details that carry consequences.
Your job is knowing the limits well enough to speak to them calmly. The score in ChatGPT may differ from the score their lender uses, the report is a month behind their bank feed, and the chatbot can't dispute an error or freeze their credit. Those three sentences make you the most useful person in your client's financial life this month.
This is also the week to talk with clients about what they're connecting and what happens to that data, which is exactly what the exercise below helps you do.
Advisors Spend Nine Hours a Week on Admin and Three Finding New Clients. Envestnet Is Spending $35 Million to Change That.
Three weeks ago, this cheat sheet covered Envestnet's announcement of a $35 million surge investment in its advisor technology. The first results are now rolling out to advisors, introduced at Future Proof earlier this month as part of the company's third technology release of 2026.
The release includes a redesigned Wealth Management Platform interface, with a customizable dashboard, configurable widgets and a self-service portal for managing unified portfolios. Envestnet says the rebuild is the foundation it needs to deliver AI features across the whole platform, which is a plain way of saying the AI is coming and the interface had to be rebuilt first.
The company also pointed at the reason all this money is moving, a Cerulli finding that out of a typical 40-hour week, advisors spend about nine hours on administrative tasks and just three hours prospecting for new clients. Nine hours on paperwork, three on growth. That ratio explains a lot about how growing a practice feels.
Why you should care: The follow-through matters more than the announcement did. Big platform promises often stall, and this one went from announcement to release in under two weeks, which tells you the money behind advisor AI is real and moving fast.
But the Cerulli ratio is the part to take personally, because you don't need Envestnet to fix yours. Write down where your own 40 hours went last week and circle the biggest administrative block. That circle is your next AI project, whether it's meeting notes, client reporting or follow-up emails.
Every hour you move from the nine-hour pile to the three-hour pile is growth you didn't have to buy.
Watt, the Amplify for Advisors Agent
See Watt solve this.
Paste your website once, and Watt drafts a week of posts with graphics in your voice and brand, ready for you to approve.
This week · ready for your review
Open now. Watt drafts a week of posts, emails and graphics in your voice, with no client data needed.
Start your 7-day free trialONE THING TO TRY THIS WEEK
Write the one-page AI guide your clients actually need. The research says they trust what AI tells them about money, and as of Monday they can hand it their credit report too. Give them the guardrails in your voice. About 20 minutes.
Step 1. Open Claude or ChatGPT and paste this.
I'm a financial advisor creating a one-page guide for my clients called Using AI for Money Questions. Interview me one question at a time, waiting for each answer, covering who my clients are, the money questions they most often bring me, how I naturally talk to them, and any rules my firm has about client communications.
Then write a one-page guide in my voice with three short sections. First, what AI tools are genuinely good at, such as explaining financial terms, organizing thoughts and preparing questions for our meetings. Second, where they often get things wrong, such as specific tax calculations, account rules, deadlines and anything with exact dollar amounts. Third, three simple habits, ask AI to explain rather than decide, never type account numbers or personal identifying information into a chat, and bring anything involving a real decision to me before acting on it.
Keep the whole guide educational, recommend no products or strategies, promise no outcomes, include no client names or identifying information, and mark it as a draft for my review and my firm's compliance process before anything goes to a client.
Step 2. Edit the draft until it sounds like something you'd say across a kitchen table, then run it through whatever review process your firm uses for client communications.
Step 3. Send it to a few clients or bring it to your next review meetings and ask a simple question, what have you been asking AI about money lately? After this week, some of them may have connected real accounts, so the conversation matters more than it did a month ago.
Your clients are already using AI for money questions. Now you're in that conversation instead of outside it.
Sam is the creator of Amplify for Advisors and Watt, the Amplify for Advisors Agent. A former RIA owner whose work has been featured in Forbes, NerdWallet, Nasdaq, and more, he teaches financial advisors to use AI in their own voice while staying compliant.