A CFP at a boutique RIA opens ChatGPT on a Wednesday afternoon. He has a client review memo due Thursday. He pastes the client's account summary and types: "Help me draft the review notes." He hits send.
That memo contains a real client name, real account values, and a real Social Security number. FINRA is already examining firms on exactly this.
The door to AI is open for financial professionals. The question is whether you walk through it correctly or whether you walk through it and straight into an exam finding. Most advisors who want to use AI have nobody showing them where the line is. Here it is.
According to AICPA and CIMA's December 2025 survey of 1,446 senior finance and accounting professionals, only 8% of finance organizations feel very well prepared for AI adoption, and 56% cite generative AI as their most prominent skills gap. The majority of financial advisors want to use these tools. Most don't have a clear path in that doesn't feel like it could blow up on them. The safe entry point exists. Here's what it looks like.
The Three Uses With No Compliance Risk
The fastest way to start using AI compliantly is to start with use cases that don't require client data to enter any AI system at all.
First: Draft template client communications with placeholder data.
Write prompts that describe the situation without naming the client. "Draft a letter to a 62-year-old client approaching Medicare eligibility who holds a significant portion of their retirement savings in pre-tax accounts, explaining the case for considering Roth conversions in the years before RMDs begin."
That prompt has no client name, no account number, no real identifying information. The output gives you a polished first draft. You fill in the actual details yourself before sending.
Advisors underestimate how much of their communication work is structurally repetitive, the same situations, different clients. Templates built with AI can cut drafting time on those communications significantly.
Second: Prepare for meetings by asking AI to explain complex concepts in plain English.
Before a client meeting on a topic you haven't lived with recently, a qualified opportunity zone investment, the rules on inherited IRA distributions post-SECURE 2.0, how a charitable remainder trust interacts with estate tax exposure, use AI to get a plain-English orientation.
No client data in the prompt. Just you getting up to speed faster. This is low-risk, high-value, and underused.
Third: Use Perplexity for investment research synthesis.
Perplexity is a search-AI hybrid that cites its sources, actual links to actual documents. For pulling together recent research on a sector, understanding a regulatory change, or getting a fast summary of an asset class you're evaluating, it's better than a standard Google search because it synthesizes rather than just lists.
The citation model matters. You can verify what it tells you, which is exactly the diligence posture financial professionals need from any AI tool.
What to Avoid Until Your Firm Has a Framework
Until your firm has an approved AI tool with an appropriate data processing agreement, one rule covers most of the risk: no actual client data into any AI system.
No names. No account values. No Social Security numbers. No financial plans. No portfolio details tied to a real person.
This isn't indefinitely sustainable, the efficiency case for AI in financial planning is real and the industry is moving. But starting with compliant, no-data-exposure use cases while your firm builds its AI policy is the right sequence. It gets you the skills and the workflows. It doesn't put you in front of a regulator explaining why you used an unapproved tool with client information.
The FINRA Reality: This Is Already Being Examined
FINRA's 2025 annual report added an entire section on AI risks. This is new. It did not appear in prior annual reports in this form. The section covers firms' obligations around supervising AI-assisted activities, the need for governance frameworks governing AI tool use, and the risks of AI-generated content in communications with customers.
FINRA has already examined firms on their AI governance. This is not regulatory speculation, it's happening in the field.
What FINRA is looking for maps reasonably well to what you'd expect: is there a written policy governing AI use? Are communications generated with AI assistance being supervised and reviewed? Is the firm able to demonstrate that client data is being handled appropriately?
If you're using AI tools without documented policies, you have an exam gap. If your firm has banned AI tools and you're using them anyway, you have a worse one.
The smart path is to get ahead of this: advocate for a firm-level AI policy that identifies approved tools, defines acceptable use cases, and establishes supervision protocols. That's better for advisors who want to use AI and better for the firm when examiners ask.
Where the Time Math Actually Works
Advisors who have built compliant AI workflows are reporting numbers that should get your attention.
Zocks, a tool built specifically for financial advisor meeting notes and CRM updates, reports in its own data that advisors cut meeting administration time by up to 90%. That's vendor-reported, verify against your own workflow, but the direction is consistent with what advisors using similar tools describe. That's not a rounding error. If you spend two hours on meeting prep, notes, and follow-up tasks per client meeting, a 90% reduction is an hour and forty-five minutes back per meeting.
Advisors using AI for report drafting and client communication are reporting comparable time savings on written output. Not because the AI is doing their thinking, but because getting a structured draft in 30 seconds and editing it to match your voice is faster than writing from a blank page every time.
The ROI is real. The compliance pathway to it just requires care about which tools you use and in what configuration.
Why Claude for Financial Professionals
The guide linked below centers on Claude, built by Anthropic. I recommend it specifically for financial professionals because of how it's designed.
Claude is built for precision and careful reasoning. It's more willing to say "I'm not certain about this" than most other AI tools, which is exactly what you want when you're working on financial concepts with real stakes. It's better at maintaining accuracy on regulatory and technical details than the average LLM, and it declines to fabricate when it doesn't know something.
For the compliant use cases, concept explanation, communication templates, meeting prep, Claude consistently produces more accurate, better-structured financial language than the default ChatGPT output. That matters when you're working in a field where precision is a professional obligation.
Consumer Claude has the same data cautions as any other consumer AI tool. But for no-client-data use cases, it's a meaningful step up in output quality for financial work.
Run this now before you close this tab:
Open Claude or ChatGPT (free tier is fine). Paste this prompt, no client data required:
"Draft a letter to a 62-year-old client approaching Medicare eligibility who holds a significant portion of their retirement savings in pre-tax accounts, explaining the case for considering Roth conversions in the years before RMDs begin. Keep the tone clear and precise."
Review the draft. That's the no-compliance-risk use case in action. No name, no account number, no identifying information, just a polished first draft you edit and personalize before sending. That's your starting point.
This article covers the safe entry point and the three no-compliance-risk uses. The CFP in the opening story got lucky. His firm didn't catch it before he did.
What it doesn't cover is what happens after you've gotten started. How do you write the written AI policy FINRA is already examining for? What supervision and review obligations attach to AI-assisted client communications? What's your exposure when you're using compliant tools but your firm has no documented governance framework?
That layer, the one where exam findings actually land, isn't in this article. It's in the guide. Financial professionals who have built AI into their practice without a compliance event didn't figure this out on the fly. They had a framework before they needed it.
You can create now.
The door to AI is open for financial professionals who know how to use it compliantly. The guide is the framework.
For CPAs, bookkeepers, financial advisors, and wealth managers, the data hygiene framework, FINRA governance checklist, and prompt library built for regulated work.
See the guide →Every AI guide for financial advisory and regulated practice work. The compliance layer is already built in.
See the full collection →If you'd rather hire the employee than read the guide, we built that. Meet the team →
Mark Reeves writes practical guides on using AI tools in real work. AI Field Guide covers Claude, ChatGPT, Perplexity, Cursor, and GEO, for business owners, solopreneurs, marketers, authors, and clinicians. See all 36 guides.