AI agents for financial advisors: AI for financial planners that researches prospects, drafts for review and keeps the CRM current.

What the rules actually allow

The framework is less mysterious than it feels. Under the SEC Marketing Rule, Advisers Act Rule 206(4)-1, an adviser is responsible for its advertisements. The rule does not ask who drafted the copy, which means an AI-written prospecting email carries exactly the same obligations as one an advisor wrote by hand. For representatives of broker-dealers, FINRA Rule 2210 requires that retail communications go through principal review and approval before use. And the books-and-records requirements under Advisers Act Rule 204-2 mean advertisements and client communications have to be retained, so a draft that goes out is a record, whatever produced it.

Put together, those three points define the workflow rather than forbidding it. The agent can research, draft, prepare and log. A qualified human reviews and approves anything client-facing. Everything sent is retained the same way it always was. What the agent must never do is give investment advice, make a recommendation, or send to a client without review, and it should never touch performance claims or anything resembling a testimonial. Treat it as a very fast marketing and operations assistant who is not licensed, because that is the correct mental model. This is general information and not legal or compliance advice, so run your workflow past your own CCO or counsel before you turn it on.

The jobs that pay off first

Meeting preparation is usually the fastest return. An advisor with a full review calendar spends real hours reconstructing context: what was discussed, what was promised, what changed in the client situation, what needs asking this time. An agent assembles that brief from the record before each meeting, in the same format every time, so preparation becomes reading for five minutes rather than digging for thirty.

Second is CRM hygiene, which every practice knows it is bad at. Notes get written when there is time, tasks get logged when someone remembers, and the pipeline slowly stops matching reality. An agent turns each meeting into properly logged activity and next steps, every time, without motivation being a factor. That is worth more than it sounds, because the quality of everything downstream, reporting, follow-up, compliance documentation, depends on the record being accurate. The CRM use case covers how that runs.

Third is prospect and centre-of-influence research. Before a meeting with a CPA, an attorney or a business owner, an agent produces a short brief on the firm, the person and the public facts, so the conversation starts informed. That is research, not advice, and it carries no marketing rule exposure at all as long as it stays internal.

How the review step works in practice

The habit that makes this safe is queueing rather than sending. The agent produces a draft and puts it in front of the person who is allowed to approve it. Nothing has a path to a client that skips that step. Practices that scale this successfully write it down as a policy, name who approves what, and keep the record of the approval alongside the communication.

The related discipline is keeping obligations and controls visible rather than remembered. Small practices tend to run compliance from one person's head and a spreadsheet, which works until it does not, and firms that grow past that usually [keep their obligations and controls mapped in one place](https://complies.ai) so a review step is a documented control rather than a habit. Whatever tooling you use, the principle is the same: if an examiner asks how AI-assisted communications get approved and retained, you want a written answer.

What it costs, and what it does not replace

For a solo advisor or a small RIA, the alternative under consideration is usually a client service associate: a full salary plus benefits for a role that is largely administrative, plus the months of ramp on your CRM, your custodian paperwork and your clients. That hire makes sense when you need a person on the phone with clients and handling service work. It makes less sense when what you actually needed was consistent research, drafting and record-keeping.

WorkAgent is $149 a month, flat, with no per-seat fee, and it does not give advice, does not talk to clients unsupervised and does not replace the advisor relationship, which is the entire product you sell. It absorbs the preparation and paperwork volume around that relationship. If you want the wider view of what an agent handles across a practice, the AI executive assistant page covers the same jobs without the advisory framing, and the AI agent pricing page has verified prices across the category.