Documenting Suitability Determinations With Agentic AI: Where Humans Still Decide

A suitability determination is a judgment, not an extraction. Picture an advisor at the moment that judgment crystallizes: a client's risk-profile worksheet open in one window, the household's allocation chart in another, and a planning-software cash-flow projection on a third screen. The advisor weighs the recommendation against three or four things at once - profile, capacity, alternatives, conflicts - and the decision lands in a single sentence the advisor will type into the meeting record. Agentic AI can do a lot of useful work around that moment. It cannot replace the moment itself, and a meeting-notes product that pretends otherwise is one that will get the firm in trouble.

Picture also the shape of the meeting note this article is about: a structured record with sections for recommendations, alternatives, and a clearly-bordered "Suitability rationale" field showing an editor sidebar with a small lock icon next to it. The lock icon means the AI did not write that field. That single design choice is the cleanest line we have seen between AI-assisted documentation that holds up under examination and AI-assisted documentation that does not.

What suitability actually is

Suitability has multiple regulatory homes. Per FINRA Rule 2111, broker-dealers must have a reasonable basis to believe a recommended transaction or strategy is suitable for the customer based on the customer's investment profile. Per the SEC's Regulation Best Interest Care Obligation, broker-dealers must exercise reasonable diligence, care, and skill to have a reasonable basis to believe a recommendation is in the retail customer's best interest. RIAs operate under the Advisers Act fiduciary standard, which absorbs suitability into a broader duty to act in the client's best interest given the engagement.

What all three frameworks share is the structural shape of the determination: profile in, recommendation out, with a documented rationale connecting them. The rationale is the suitability determination. The profile and recommendation are inputs and outputs, both of which AI can populate. The rationale is the human judgment.

Where the AI legitimately helps

An agentic note assistant can do real work around the determination without making it. The supporting pieces it can produce:

• A current snapshot of the investment profile fields, with a flag if any field has aged past the firm's refresh cadence.
• A summary of the recommendation made in the meeting, with citation back to the moment in the conversation.
• A list of alternatives the conversation considered, with the reason each was set aside.
• A note of any conflicts disclosed during the meeting and how the client acknowledged them.
• A surfacing of any contradiction between the profile on file and statements the client made in the meeting (a job change, an inheritance, a health event).

That is a lot of work. None of it is the suitability rationale itself.

Where the human stays in

The suitability rationale - two to four sentences explaining why this recommendation fits this profile right now - is the field that should remain advisor-authored. Three reasons hold consistently across our pilot:

1. Defensive posture under examination. A record where the suitability narrative is human-authored and timestamped separately from the AI-drafted body cannot be challenged as "AI-generated suitability."
2. Fiduciary line-drawing. The duty to act in the client's best interest is not delegable to a tool. Keeping the rationale field human-authored keeps that line bright.
3. Quality. The advisors in our pilot wrote noticeably better rationales when they wrote them themselves than when they edited an AI-drafted version. The AI version was reasonable; the advisor version reflected information the model did not have access to (the prior conversations, the household dynamics, the planning trajectory).

The lock-icon principle. Every field in a meeting note should be tagged either AI-drafted or human-authored, and the suitability rationale should always be the latter. If the product cannot enforce that distinction in the record, the firm has not yet earned the documentation-time savings.

The contradiction-flagging pattern

The most useful AI behavior we have shipped in the suitability area is also the simplest: flag, do not decide. When a meeting transcript contains a statement that contradicts the on-file profile - the client mentions a job change, a major liquidity event, a divorce - the assistant surfaces the contradiction in the draft and waits. The advisor decides whether to update the profile, and the update is logged as its own record event with its own timestamp. Across our early advisor cohort, this pattern surfaced a profile-aging cue in roughly one in seven meetings, with most of those cues being ones the advisor would not have flagged unaided.

What about Reg BI's documentation requirement

Reg BI requires a record sufficient to demonstrate the firm had a reasonable basis to believe the recommendation was in the customer's best interest. "Sufficient" is a working standard, not a fixed format. A record where the recommendation, rationale, alternatives, and conflicts are all named - with the rationale advisor-authored - is the cleanest version of "sufficient" we have seen survive examination cycles. Across our pilot, the firms that adopted this division of labor produced records that examiners moved through quickly; the firms that let the AI write the rationale produced records that triggered follow-up requests at noticeably higher rates.

Closing the loop

The recommendation here is structural, not aspirational. Configure the documentation tool to leave the suitability rationale field empty in the AI draft. Train advisors to fill it in two to four sentences, in plain English, naming the profile attributes the recommendation aligns to. Audit the field quarterly. The combined effect is a documentation corpus where the time savings are real, the AI's role is bounded, and the fiduciary judgment is exactly where regulators expect it to be.

Source notes

• FINRA Rule 2111 suitability obligations.
• SEC Regulation Best Interest Care Obligation guidance and staff Q&A.
• SEC Advisers Act fiduciary standard and Investment Adviser Information Sheet.
• NIST AI Risk Management Framework, Generative AI Profile, on human-in-the-loop patterns.
• IAA 2026 Investment Adviser Industry Snapshot on documentation practices among independent RIAs.