Disclosing AI in your ADV isn't enough

If AI helps write a client email or fill out a form, your firm should be able to show where the facts came from, who checked the work and what went out. A risk disclosure in your ADV doesn't show any of that. For SEC-registered advisers, Rule 204-2 covers required records and Rule 206(4)-7 requires written compliance policies.

Every claim needs a source

An adviser asks an AI agent to prepare an annual review. The summary says the clients decided to change their 529 beneficiary, but the meeting note may say they were only considering it. The adviser should be able to click that sentence and see the source before raising it with the client. FINRA's 2026 report names summarization as the top GenAI use case at member firms and flags inaccurate outputs presented as fact.

Review before the work leaves the firm

When AI prefills a beneficiary-change form, a wrong date of birth can look perfectly plausible. "Human in the loop" only means something if the reviewer sees the source for each field and approves the form before it goes to the custodian.

The model is one step of the workflow

Whether a firm uses a general AI workspace, a platform built for advisers or its own build, the steps are the same. What changes is who runs them and who keeps the evidence. Zero data retention limits what a model provider keeps. It doesn't decide what the adviser has to keep.

A quick test

Pick one AI-assisted client task from last week and trace it from the source document to the final approval. If assembling that takes hours, the check won't happen consistently, and no ADV sentence fixes that.

This is a short version of an issue of The Prodocol, our newsletter on LinkedIn. Read the full issue or subscribe to The Prodocol.