Do I need a paid AI account to do this work?

Yes for any production RIA workflow. The free tiers of ChatGPT, Claude, and Gemini do not give you a Data Processing Addendum, do not let you turn off training on your inputs by default, and do not produce the kind of audit log a CCO or SEC examiner expects to see. ChatGPT Team or Enterprise, Claude Team or Enterprise, and Microsoft Copilot for Microsoft 365 with the right configuration are the realistic starting tiers for an RIA. Pricing runs $25 to $60 per seat per month depending on tier. For a five-IAR firm that is roughly $1,500 to $3,600 a year, which is less than a single hour of partner time saved per week.

Is ChatGPT or Claude SEC-compliant for client data?

The tools themselves are not SEC-compliant or non-compliant. Compliance lives in the workflow around them. The SEC Marketing Rule and the Custody Rule do not name specific software. What they care about is whether client communications are accurate, fair and balanced, archived correctly, and free of prohibited performance claims. The practical rule: use the Enterprise or Team tier with the no-training option enabled, document the workflow in your written policies and procedures, log every AI-assisted client communication in your archive of record (Smarsh, Global Relay, MyComplianceOffice), and have your CCO sign off on the prompt templates the firm uses. Do not paste account numbers, SSNs, or full account values into the consumer tier.

Will the AI-generated meeting prep sound generic and miss what makes my client relationship unique?

Only if you prompt it that way. The IARs who get useful output give the AI specific structure: the client's stated goals from the last review, the planning topics that came up in the last two meetings, the household composition, the target asset allocation, and the things the client cares about that have nothing to do with the portfolio. Generic prompts produce generic output. Specific prompts produce a meeting brief that lets you walk in sounding like you remembered everything they told you in March. The trick is treating the AI as a reader of your CRM notes, not as a content generator.

How do I share AI-generated meeting briefs with my paraplanner or junior advisor if they do not have an AI account?

Export the brief as a PDF or paste it into your CRM activity record. Both Redtail and Wealthbox let you attach a file or write a long-form note to the contact record. The brief lives in your existing system, the team accesses it the way they already access meeting notes, and the AI account stays a tool the prep person uses, not a system everyone has to log into. For firms running Salesforce Financial Services Cloud, the paraplanner can pull the brief from the activity timeline the same way they would pull any other document. The AI tool is upstream of the CRM, not parallel to it.

What if my broker-dealer or CCO blocks AI tools entirely?

Three things to do. First, ask specifically what is blocked and why. The answer is usually consumer-tier tools without a DPA, which is the right call. Enterprise tiers with proper data agreements are usually negotiable. Second, propose a small pilot with one specific use case (meeting prep is the easiest sell because it is internal-facing and does not produce client communications directly). Third, document everything: the tool, the data, the retention, the human review step. CCOs say no to vagueness. They say yes to a written workflow. If the firm is registered with FINRA in addition to the SEC, you also need to address how the workflow logs into your 17a-4 retention system before the pilot starts.

Can my clients use AI tools to ask questions about their accounts directly?

Not through your firm. Any direct AI-to-client interaction about specific accounts crosses several lines: the SEC Marketing Rule (any communication that could be construed as advertising), fiduciary duty (the AI cannot exercise discretion or give personalized advice), and FINRA archiving requirements if you are dual-registered. The supportable answer for clients who ask is that the firm uses AI internally to make IARs better prepared, faster, and more thorough, but every client communication still goes through a licensed advisor. That is the truthful and defensible position. Clients using ChatGPT on their own to ask general financial questions is fine. The firm letting AI talk directly to clients is not.

I am not technical and our firm runs on Redtail and Schwab. Can I actually do this?

Yes. The workflow described in this guide is a copy-paste workflow, not an integration build. You pull the relevant data from Redtail (last meeting notes, household summary, planning topics), paste it into a Claude or ChatGPT Team conversation with a structured prompt, and get back a meeting brief you read and edit. No API work, no custom code, no IT help. The advanced version of this involves a Redtail API integration or a tool like Catchlight or Jump that pulls from Redtail directly, but the manual version works fine for a five-to-fifteen-IAR firm and produces the same quality of brief. Start manual. Automate later if the volume justifies it.

How do I actually log AI-assisted prep work in our 17a-4 archive?

If the AI output ends up in a client communication, the communication itself is what gets archived, not the prompt or the AI session. If the AI output is purely internal prep (a meeting brief the IAR reads before walking in), most firms treat it the way they treat any other internal work product: store it in the CRM activity record, retain it under the firm's records retention policy, but not necessarily in the 17a-4 archive of record. The line is whether the document is a client communication or an internal staff document. Confirm the treatment with your CCO and put it in writing in the firm's WSPs. Smarsh, Global Relay, and Hearsay all now offer connectors that capture AI session logs if you want belt-and-suspenders coverage, but for most RIAs the activity-record approach is enough as long as the policy is documented.

How Do RIAs Use AI for Client Meeting Prep Without Violating SEC Marketing Rule?

Most RIAs I talk to spend two to four hours preparing for each annual review. Pulling performance data from the custodian. Re-reading the planning notes from last September. Building a one-page client summary. Drafting agenda talking points. Looking up whatever the client mentioned last time about the daughter's wedding or the rental property in Sarasota. By the time the IAR walks in, three hours of billable time is gone and the brief is half-remembered.

Multiply that across a 120-household book and the math gets ugly. A solo IAR with 80 households loses roughly 200 hours a year to meeting prep. A four-IAR firm with 400 households loses 1,000 hours. Those are hours that cannot be billed and cannot be reclaimed by hiring more paraplanners without compressing margins.

AI does not solve this by replacing the IAR's judgment. It solves it by handling the assembly work that comes before judgment: pulling the client context together, structuring the brief, surfacing the questions the client is likely to ask, and drafting the talking points the IAR edits in five minutes instead of writing in fifty. The IAR still owns the meeting. The AI just means the IAR walks in prepared.

This guide walks through the meeting prep workflow that holds up under SEC Marketing Rule scrutiny, FINRA 17a-4 archiving requirements, and fiduciary duty. It includes the prompt patterns that work for RIAs, the compliance line every firm needs to draw, and a starter template for your next quarterly review.

Why this matters for RIAs specifically

RIAs are uniquely under-served by current AI tooling because the consumer tools were built for marketing teams and product managers, not fiduciary advisors. The defaults are wrong. Training on inputs is on by default. Outputs are framed as confident assertions instead of caveated recommendations. The tools love adjectives the SEC has been telling RIAs to stop using since 1940.

The firms that figure out the workflow, the prompt patterns, and the compliance line get hours back per IAR per week without changing what the firm does for clients. They keep the fiduciary standard intact, keep the 17a-4 archive clean, and use AI to compress the assembly work nobody enjoys. Firms that wait usually end up either banning AI awkwardly (and watching IARs use it on personal accounts in a parallel workflow nobody can audit), or rolling out a vendor tool nobody trained on. Both are worse than a documented, deliberate workflow.

What AI for meeting prep actually does

The useful tools are general-purpose large language models (Claude, ChatGPT, Microsoft Copilot) running on Enterprise or Team tier with the right data agreement in place. They take inputs (CRM notes, custodian performance summary, planning topic list, household details) and produce a structured meeting brief, draft agenda, and talking points.

Three things make these tools different from the AI features bundled into your CRM:

They handle unstructured inputs. Paste raw meeting notes from a year ago, a 14-page financial plan, and a custodian-generated PDF and the AI reads all of it. Most CRM-bundled AI features only work on structured fields.
They produce structured output to your specification. Define the brief format once. The AI produces it the same way every time. Consistent format is auditable.
They iterate in plain English. "Move the 401(k) rollover question to the top, drop the section on the rental property, add a question about Section 199A" is a one-sentence edit.

Think of it as a paraplanner who has read everything in the file, never gets tired, and produces the brief in the exact format your firm uses every time.

Before you start

You need:

A Claude Team, ChatGPT Team or Enterprise, or Microsoft Copilot for Microsoft 365 account with training-on-inputs disabled. Pricing is $25 to $60 per seat per month.
About 45 minutes for your first session, mostly to build and refine the prompt template.
One real client review coming up in the next two weeks. The workflow is not abstract. You will be doing real prep work.
Read access to your CRM (Redtail, Wealthbox, Salesforce FSC) and your custodian's performance reporting (Schwab, Fidelity, Pershing).
Your CCO or compliance designee on a 30-minute call to review the prompt template before the first client uses it.

One thing to settle before you paste anything: the SEC Marketing Rule, FINRA 17a-4 archiving, and your fiduciary duty all shape what AI can and cannot do here. We have a dedicated section on this below. It is non-negotiable.

Material 1: The household context brief

The failure pattern most IARs fall into: walking into a review meeting having skimmed last year's notes for ten minutes on the way to the office. The client mentions something they told you eight months ago and you have to fake it. They notice. Trust takes a small hit even if nothing else goes wrong.

What to ask the AI for instead:

I have a client review meeting tomorrow with [Client Last Name] household. Below are the last three meeting notes from our CRM, the household summary, and the financial plan from 2024. Build me a one-page household context brief with these sections: (1) Who they are in 2 sentences, (2) Stated goals at last review with progress notes, (3) Planning topics that came up across the last three meetings, (4) Personal details they have shared (kids, hobbies, life events, anything not portfolio-related), (5) Three open threads they are likely to ask about. Keep it factual. Do not infer or speculate. Where the notes are silent, mark the section blank rather than fill it in.

The prompt is doing several things at once. It specifies the input (CRM notes, household summary, plan), the output structure (five named sections), and the constraint that matters most for an RIA: do not infer or speculate. AI tools will happily make up details to fill a section. The "mark it blank" instruction is what stops that.

For smaller households (sub-$1M, simpler plans), you can compress this to a half-page brief with the same structure. For larger households (multi-generational, trust structures, business interests), expand with sections for entity structure and key advisors (CPA, attorney, insurance broker). The structure stays the same. The depth scales.

Material 2: The performance review draft

The single most regulated thing in any RIA review meeting is how you talk about performance. The Marketing Rule rewrote what is and is not allowed in performance presentation, and what was acceptable in 2018 is not acceptable in 2026. Hypothetical performance has new disclosure requirements. Net-of-fees presentation is mandatory in most contexts. Predecessor performance has its own rules.

The failure pattern: an IAR builds a performance summary in Excel, ports it into a Word doc, and the disclosures are out of date by 18 months because nobody on the team owns keeping them current.

What to ask the AI for instead:

Below is the YTD and trailing 1, 3, and 5-year performance for [Client Household]. The numbers are pulled from Schwab Performance Reporting on [date]. Draft a performance review section for my review brief that: (1) States the time-weighted return net of fees for each period in a clean table, (2) Compares to the household's stated benchmark from the IPS, (3) Frames the discussion in terms of progress against goals, not market commentary, (4) Surfaces any account-level result that diverges meaningfully from the benchmark and notes it as a discussion point, (5) Includes the standard performance disclosures the firm uses, which I will paste below. Do not characterize performance as good or bad. Do not project forward. Do not use the words strong, solid, healthy, or any other adjective the SEC has flagged in Marketing Rule guidance.

The prompt move that matters: the explicit ban on adjectives. AI tools default to characterizing data ("a strong year for U.S. large cap") because that is how human analysts write. RIAs cannot. The Marketing Rule's prohibition on cherry-picking and on misleading framing means the safest performance presentation is the one that states the numbers, names the benchmark, and lets the IAR explain the context in person.

For a discretionary firm, add a sentence about the IPS-defined rebalancing tolerance and whether any positions hit the rebalance band. For a non-discretionary firm, add a section on account changes the client requested versus account changes the IAR recommended that the client declined, because that history matters for future-meeting context.

Material 3: The planning topic queue

Most client reviews are organized around the portfolio. The planning topics get squeezed into the last 15 minutes, the IAR runs out of time, and three things that should have been handled get pushed to the next meeting. Six months later they are still pushed.

The AI move here is to build a planning topic queue that the IAR walks in with, sorted by urgency.

Read the household context brief, the financial plan from 2024, and the last three meeting notes. Build me a planning topic queue for [Client Household] sorted into three buckets: (1) Time-sensitive, things that have a hard deadline in the next 90 days (RMDs, tax filings, beneficiary updates after a life event), (2) Worth raising at this meeting, things that have come up in the notes but never been resolved, (3) Background topics, things to keep an eye on but not necessarily address this meeting. For each item, write one sentence on why it matters and one sentence on what the IAR should ask or recommend. Do not give specific tax or legal advice. Frame anything tax-adjacent as something to confirm with the client's CPA.

The Circular 230 line matters here even though Circular 230 is a CPA framework, not an SEC framework. Any tax-adjacent recommendation an IAR makes can edge into territory the IAR is not licensed to give advice on. The prompt move "frame anything tax-adjacent as something to confirm with the client's CPA" keeps the brief on the right side of the line and protects the IAR if the client takes the meeting brief as definitive guidance later.

For households with trust structures or business interests, add a section for legal-adjacent items framed the same way: confirm with the client's attorney. The IAR's job is to surface the question, not to answer it.

Material 4: The agenda and talking points

The failure pattern: a typed agenda that says "Review performance, planning update, Q&A, next steps" for every meeting, every quarter, regardless of what is actually going on in the household.

The AI version is a custom agenda for this household, this meeting, with talking points written in the IAR's voice.

Build me a meeting agenda for the [Client Household] review tomorrow. Five sections, 60 minutes total. Allocate time based on the planning topic queue and the household context brief. For each section, write three talking points or questions in my voice (I will paste a sample of how I write below for tone reference). Talking points should be open-ended where the IAR is gathering information and direct where the IAR is recommending action. Mark anywhere the IAR is making a recommendation that requires fiduciary judgment with [JUDGMENT] so the IAR knows that bullet is not a script to read but a decision to make.

The "voice match" prompt is the one most IARs skip and most regret skipping. AI tools default to a generic financial-services voice. If you paste two paragraphs of how you actually write, the agenda comes out sounding like you, not like a stock advisor newsletter.

The "[JUDGMENT]" marker is the fiduciary compliance move. It draws an explicit line in the brief between assembled context (which the AI handles fine) and judgment calls (which only the IAR can make). Anyone reviewing the brief, including a CCO doing a spot audit, can see exactly which bullets are AI-assembled and which require IAR sign-off.

Material 5: The post-meeting follow-up draft

The meeting ends. The IAR has fifteen minutes before the next call. The follow-up email gets dashed off, the action items get half-captured, and the meeting note gets written from memory three days later when the IAR finally sits down to do CRM hygiene. By then the details have softened.

AI handles this part better than the prep part if the workflow is set up correctly. After the meeting, the IAR dictates a five-minute voice memo: what was discussed, what was decided, what is next. The AI takes the voice memo and produces three things in parallel.

I just finished a review meeting with [Client Household]. Below is a voice memo transcription of what we discussed. Produce three deliverables: (1) A clean meeting note for the CRM activity record, factual and structured, with sections for Discussed, Decided, Action Items, and Next Steps. (2) A draft client follow-up email summarizing what we agreed and confirming the action items, in my voice (sample below). (3) A list of internal action items for me, the paraplanner, and operations, with each item marked by owner. Do not invent details that are not in the transcription. If the transcription is silent on something the client mentioned earlier in the relationship, mark it for me to confirm rather than fill it in.

The meeting note goes into the CRM activity record. The follow-up email gets reviewed by the IAR, edited in two minutes, and sent. The internal action items go to whatever task system the firm runs (Asana, ClickUp, the CRM's task module). All three from one voice memo, all three in the same structure every time.

For compliance review, the meeting note in the CRM is the document of record. The AI helped produce it, but the IAR signs off on the final version. That is the line. The AI is a drafting tool. The IAR owns the document.

Material 6: The quarterly client communication

Quarterly client letters and market commentary are the highest-risk area for AI in an RIA workflow because they are unambiguously client communications under the Marketing Rule and subject to FINRA 17a-4 retention if the firm is dual-registered. The temptation is high (a quarterly letter is a clear use case) and the compliance exposure is also high.

Below is [the firm's quarterly market commentary template] and [the firm's house view as approved by the IC for Q3 2026]. Draft a 700-word quarterly client letter for the firm's signature in our standard voice. Do not introduce market views beyond the IC house view. Do not predict or project specific market levels. Do not characterize the firm's portfolios as outperforming or underperforming. Use the standard disclosures, which I will paste below. The letter is a draft. The CCO will review and finalize before it goes out.

The constraint that protects the firm: the IC house view is the only source of forward-looking commentary. The AI cannot generate forecasts. The CCO review is the final gate. The 17a-4 archive captures the version that goes out, not the AI draft, but the firm's WSPs should explicitly cover how AI-assisted drafts fit into the review workflow. For firms without a formal IC, replace "IC house view" with "the firm's approved commentary as signed off by the principal."

The RIA-specific prompts that actually work

After watching IARs use AI for meeting prep through a couple of cycles, the difference between a usable brief and a generic-sounding one comes down to four prompt moves.

Specify the audience and the meeting type. "Annual review for a $4.2M household, retired couple, age 67 and 65, two adult children, one with a special needs trust" lands very differently than "client review." The AI calibrates depth and tone to the household.

Specify the constraint that actually matters. For an RIA, the constraints are usually: do not project performance, do not cross into specific tax or legal advice, do not characterize performance with adjectives the Marketing Rule restricts. State them explicitly. The AI honors explicit constraints. It guesses badly at implicit ones.

Specify the firm voice and document format. Paste a sample of an existing brief or email in the firm's actual voice. The AI matches it. Without the sample you get the AI's default voice, which is generic financial-services blandness.

Specify what stays static and what changes. For recurring meeting types (quarterly review, annual review, financial plan refresh), tell the AI what is fixed (firm disclosures, IPS reference, standard agenda structure) and what is variable (this household's planning topics, this household's performance, this household's open threads). This is the move that makes the workflow reusable instead of one-off.

The SEC Marketing Rule, FINRA 17a-4, and Circular 230 non-negotiables

This section is short because the rules are simple, but it is the most important section in this guide.

Do not put any of the following into the consumer tier of any AI tool:

Client names, account numbers, or SSNs
Specific account values or holdings tied to a named client
Beneficiary information or family financial details from the financial plan
Trust documents, estate plans, or insurance policy details
Tax returns or any document with the client's TIN visible
Anything that could identify a client by name or account

Use AI for templates, workflow structure, prompt patterns, and household-context briefs that you build on the Team or Enterprise tier with a Data Processing Addendum signed and training on your inputs disabled. For client communications that go out, the AI draft is internal work product. The version that goes out gets reviewed, edited, and signed off by the IAR or CCO inside the firm's normal review process and gets captured in the 17a-4 archive of record (Smarsh, Global Relay, Hearsay, MyComplianceOffice) the same way any other client communication does.

For RIAs that are also dual-registered or that do tax-adjacent work, Circular 230 limits what the IAR can say about tax matters. Frame anything tax-adjacent as something the client should confirm with their CPA. Do not let the AI generate specific tax advice. The line is the same whether the IAR or the AI wrote it.

If your firm has signed an Anthropic Business agreement, an OpenAI Enterprise agreement, or a Microsoft 365 Copilot agreement with a Data Processing Addendum, the rules can be different. Ask your CCO and your IT director what is covered. Document the workflow in your written supervisory procedures. Do not assume.

When NOT to use AI for client meeting prep

AI is a generalist tool. It will not be the right answer for every part of an RIA's prep workflow.

Skip it for:

Anything that requires fiduciary judgment without IAR review. Specific allocation recommendations, suitability decisions, fund selection, or risk tolerance changes. The AI can surface the question. The IAR makes the call. Always.
Direct client-facing AI interactions. Letting the AI draft an email is fine. Letting the AI talk directly to clients (chatbots, AI-driven account portals that answer specific account questions) crosses several lines: Marketing Rule, fiduciary duty, supervision requirements. Those tools require a different compliance posture and are out of scope for the workflow in this guide.
Performance projections or hypothetical returns. The Marketing Rule's hypothetical performance rule is one of the strictest provisions in modern RIA compliance. AI is too willing to generate forward-looking numbers. Do not let it.
Tax-specific advice or legal-specific advice. Even if the IAR holds a CPA or JD personally, the firm's RIA registration does not cover tax preparation or legal advice. Frame, do not advise.

A simple rule: AI is an unfair advantage on the 80% of meeting prep where the work is assembly. Trust the IAR's judgment and the firm's compliance review for the 20% where the document or the decision has fiduciary or regulatory weight.

The quick-start template

Here is the prompt scaffold that works across most RIA meeting-prep use cases. Copy it, fill in the brackets, paste into Claude Team or ChatGPT Team.

Build me a [meeting brief / performance review section / planning topic queue / agenda / follow-up draft] for my client meeting with [Client Household] on [date].

Inputs: [paste CRM notes, household summary, performance data pulled from custodian, financial plan summary, anything else relevant].

Output structure: [list 3 to 6 named sections with one sentence on each].

Voice and tone: [paste 2-3 paragraphs of how the IAR or firm actually writes].

Constraints: Do not project performance. Do not characterize performance with adjectives. Do not give specific tax or legal advice. Where the inputs are silent, mark the section blank rather than infer. Mark any item requiring fiduciary judgment with [JUDGMENT].

Output format: [plain text for CRM paste, formatted for printing, draft email, etc.].

For recurring meeting types, save the populated template once and reuse the structure. The household-specific inputs change. The structure does not.

Bigger wins beyond meeting prep

Once the meeting prep workflow is running, the next layer of value shows up in places that have nothing to do with individual client meetings.

Firm-wide prep template library. Build a library of approved prompt templates: annual review, quarterly review, plan refresh, prospect first meeting, retirement income transition. Each gets CCO sign-off once. Every IAR uses it. Compliance review happens at the template level, not at every meeting.

Onboarding automation. New client onboarding is paperwork-heavy. AI handles the first-pass document review (plan intake, statement consolidation, planning topic identification) faster than any process change. The IAR walks into the second meeting with a real plan instead of a clean slate.

Internal knowledge base. The firm's institutional memory lives in partner heads and a few PDFs. AI can take the firm's existing IC minutes, planning whitepapers, and meeting notes and make them searchable. New IARs ramp faster. Senior IARs answer fewer repetitive questions.

Compliance review acceleration. AI can flag the items that need CCO attention (Marketing Rule trigger words, performance characterizations, tax-adjacent statements) so the CCO spends time on the high-risk items instead of reading every word of every email.

The financial services AI consulting connection

This is one tool in one category of an RIA's workflow. The bigger question for the financial services niche is structural: client expectations are shifting toward faster, more personalized communication, while compliance frameworks are tightening around how AI can be used in regulated communications. RIAs that figure out the workflow end up with IARs who get hours back per week, books that grow without proportional staffing, and a compliance posture that holds up under examination.

If your firm is wrestling with the bigger AI question, the AI Consulting in Financial Services page covers the full scope: where AI fits in RIAs, wealth management firms, mid-size CPA firms, and insurance brokerages, what the common failure modes look like in regulated practices, and what an engagement looks like when it works.

For individual IARs, start with this guide. Build one household context brief tonight, on a real client meeting next week. The whole thing takes 30 minutes. The meeting you walk into next week will go differently than the one you walked into last quarter.

Closing

The goal is not for IARs to become prompt engineers. It is for IARs to never have to assemble a meeting brief from scratch again. AI is the closest tool I have seen to that goal for RIAs specifically. It rewards specificity, respects the constraints of the regulatory frame, and gives back the hours that used to go into prep work that nobody enjoyed doing anyway.

Pick one upcoming review meeting. Build the household context brief tonight. See what 30 minutes of prompt work produces compared to your current 2-hour prep session. The case for the rest of the workflow makes itself after that.

If you want to talk about how AI fits into your RIA at the firm level, the AI Consulting in Financial Services page lays out the full picture and how an engagement works.