These Related Stories
Share this
What’s spookier than a client asking ChatGPT for financial advice? A client following that advice without realizing what’s missing.
A response can sound reassuring and include calculations, a clear recommendation, and a checklist of next steps. But who checked the assumptions? What important details were never included in the prompt? And who’s helping the client work through the consequences?
As an advisor, you know how much work happens before you make a recommendation. You ask questions, connect competing priorities, and sometimes uncover the concern a client didn’t know how to put into words. That judgment is a meaningful part of what clients pay for.
I think that’s where this conversation needs to begin. As clients explore artificial intelligence (AI) for help with money questions, we have an opportunity to make the value of professional advice clearer. We also need to help them recognize when they’re sharing information that should never have entered the chat.
Because the financial decisions and the data behind them can follow someone long after they close the window.
What might clients be asking ChatGPT about money?
To help clients recognize those risks, start with the decisions they may be bringing to a chatbot:
- “Should I pay off my student loans or invest?”
- “Can I afford to leave my job?”
- “Which investments should I choose in my 401(k)?”
- “Am I on track to retire?”
- “What should I do with an inheritance?”
- “Do I need a financial advisor?”
Questions like these show how quickly financial education can turn into a request for personal direction. A client may open the chat looking for an explanation and leave believing they have a plan.
“Explain how a Roth conversion works” asks for an explanation. “Tell me how much I should convert this year” asks for a recommendation with consequences.
A client may not recognize when they’ve crossed that threshold. As advisors, we have an opportunity to help them see it.
The scary part is how convincing an incomplete answer can sound
Imagine someone asks ChatGPT whether they can afford to buy a house. They provide their salary, savings, and the purchase price. The response arrives neatly organized, with estimated payments and an encouraging conclusion.
It feels thorough.
But what if they’re planning to leave their job? Supporting a parent? Expecting a major medical expense? What if their partner is deeply uncomfortable with the monthly payment?
A chatbot can ask follow-up questions and work through scenarios. But an answer still depends on the information available, the assumptions used, and whether the analysis is correct. The user may not know which missing details matter.
The U.S. Securities and Exchange Commission (SEC), North American Securities Administrators Association, and Financial Industry Regulatory Authority warn that AI-generated investment information can be inaccurate, incomplete, misleading, or entirely made up. Their joint investor alert encourages investors to verify underlying sources and consult a registered investment professional.
The risk doesn’t always announce itself with a flickering light and ominous music. Sometimes it looks like a reasonable answer with a very tidy table.
What does an advisor bring that a chatbot conversation doesn’t?
A client may understand diversification. Now what should change in their portfolio, considering their employer stock, tax situation, upcoming expenses, and willingness to tolerate losses?
A client may know how compound growth works. How much can they realistically save while paying for childcare and keeping a cash cushion that helps them sleep?
This is where an advisor’s work becomes concrete. You help connect decisions that clients may be considering separately. You examine assumptions, discuss competing priorities, and help them follow through. In an ongoing engagement, you revisit the plan when circumstances change.
There’s also a professional obligation behind the relationship. The SEC’s interpretation of investment advisers’ fiduciary duty describes duties of care and loyalty under the Investment Advisers Act of 1940. Asking a general-purpose chatbot to “act as my fiduciary” doesn’t establish that adviser-client relationship.
For a fee-only advisor, that’s a stronger value proposition than simply having an answer ready. It’s about explaining why a recommendation fits, what could change it, and how you’ll help the client act on it within the scope of your engagement.
Before anyone uploads a financial statement, check what’s behind the door
It’s easy to think, “The more information I provide, the better the answer.”
With financial information, that instinct needs a pause.
A tax return, account statement, or financial plan can contain far more than the numbers needed to answer a question. Names, addresses, account numbers, dependent information, and details about someone’s health or family may come along for the ride.
I’d encourage clients to start with general questions and fictional examples rather than uploading sensitive documents. Social Security numbers, passwords, verification codes, and account credentials have no place in a financial education prompt.
The privacy discussion also needs precision. Uploading information to ChatGPT does not automatically publish it online. But submitting it to an outside service means its data practices matter.
According to OpenAI’s data controls guidance, personal ChatGPT workspaces have a setting that determines whether new conversations can be used to improve the model. Turning that setting off does not delete saved chats.
Temporary Chat has additional limits on history and training, but OpenAI says temporary conversations may still be retained for up to 30 days for safety purposes. “Temporary” should not be read as “nothing is stored.”
Those controls are useful. They are not a substitute for deciding whether the information should be submitted in the first place.
Advisors need to protect their own information, too
The same care applies when we’re the ones writing the prompt.
I’d keep client information and proprietary firm material out of an unapproved personal ChatGPT account. That includes internal research, nonpublic pricing plans, acquisition discussions, confidential procedures, and planning methods your firm has developed.
You don’t need to upload your entire planning playbook to ask for ideas for organizing a checklist.
And replacing a client’s name with “Client A” is not necessarily enough. An unusual combination of employer, location, family circumstances, and financial details may still identify someone. A fictional scenario built from scratch can be a better starting point.
Business products deserve a separate review. OpenAI states that ChatGPT Business, Enterprise, and its application programming interface (API) do not use business data for model training by default. That is an important distinction from personal accounts. It does not, by itself, establish that a particular workflow meets your firm’s confidentiality, security, or compliance requirements.
Connected tools need attention as well. OpenAI’s guidance on custom GPTs explains that relevant input may be sent to third-party services when external apps or APIs are involved. Those services have their own data practices.
Before approving a workflow, I’d work through a few practical questions with the person responsible for compliance and security:
- Which account, product, and connected services will receive the information?
- What information is permitted, and what must stay out?
- Who can access the data, and how long is it retained?
- How will recommendations, calculations, and sources be verified?
- What records need to be retained under the firm’s applicable obligations?
An approved tool still needs an approved use. Permission to draft a public blog outline shouldn’t quietly become permission to upload client meeting transcripts.
Invite clients to bring the conversation out of the shadows
If a client has used ChatGPT to think through a financial decision, I’d want them to feel comfortable saying so.
Try asking:
“Have you used ChatGPT or another AI tool to explore this? What did it suggest?”
That gives you somewhere useful to begin. You can review the assumptions together, check sources, and explain what the response missed. Ask them to describe what they shared, too, so you can discuss privacy without making them feel foolish.
You could also offer a simple suggestion for future use:
“Use it to help you prepare questions for our meeting. Before you act on a recommendation or upload a financial document, let’s talk.”
That keeps curiosity welcome while clarifying the advisor’s role.
ChatGPT can help someone walk into a meeting more informed and more confident asking questions.
But when the question becomes “What should I do with my money?”, a convincing response is only the beginning. Clients deserve advice grounded in their circumstances, a clear explanation of the trade-offs, and someone to help them navigate what comes next.
Share this
Subscribe by email
Choosing Financial Performance Reporting Software That Works for Your RIA
.png?width=360&height=188&name=Your%20Best%20ROI%20Engaged%2c%20Happy%20Employees%20(Exclusive%20XYPN%20LIVE%20Content).png)
Choosing Financial Performance Reporting Software That Works for Your RIA
Curating Your Financial Planning Knowledge Diet: A Strategic Approach to Staying Informed
Curating Your Financial Planning Knowledge Diet: A Strategic Approach to Staying Informed
Email Newsletters Don’t Have to Be Boring
