AI Notetakers in Committee Meetings: What fiduciaries should know

If you serve on a committee for a retirement plan or an endowment or similar entity, you know the challenge: Meetings with important decisions, nuanced discussions, and action items all moving quickly while someone tries to capture it all by taking meeting notes, or worse, remember it all after the fact. Typed meeting notes have been the norm for years, and they’ve served their purpose. AI notetakers offer a new way to capture those meetings, but they also raise documentation risk questions your committee acting as fiduciaries should consider.

What your committee should know first

There’s no clear, industry standard yet for how AI notetaking should be handled in retirement plan committee or endowment meetings. Technology has moved faster than the guidance around it. Most of the open questions are about whether AI notetaking could create fiduciary documentation risks, by over-documenting.

For example, AI notes can capture informal comments that may be confusing out of context, such as a brainstormed question or idea a committee never formally adopted or discarded. A regulator or non-fiduciary stakeholder could mistake that for unfinished business that never received proper follow-up, and that could create a fiduciary liability. Similarly, if your committee has no specific policy for when AI notes can or should be deleted from the AI system, inconsistent retention or deletion could raise questions about whether or not notes are being selectively removed, and that can raise red flags. ERISA attorneys tell us that handwritten, typed, and AI-generated notes could all be discoverable in the event of litigation. In other words, the documentation risks are not new to AI, but AI notes add new documentation dynamics to manage.

If your committee uses its own AI notetaker

We encourage you to develop a policy that improves efficiency without increasing fiduciary liability. That includes deciding when or how long the original AI notes are kept or deleted. We’d also encourage your committee to:

• Use only tools that do not put meeting information into a learning model or the public domain
• Vet AI tools for privacy and security
• Keep confidential information out of public AI platforms

How The Advisory Group is approaching AI notes

The standards above are the same standards we apply ourselves. Starting mid-2026, The Advisory Group is using an AI notetaker in most of our client meetings, unless a client committee objects. Our goal is to give clients the full-service benefit of AI notetaking without creating liability for client committees. Because we are a fiduciary but not part of your fiduciary committee, we take our notes independently, by hand or with AI. We make sure the final notes are reviewed and tuned to capture only accurate and relevant information.

During each meeting, the AI tool listens in and generates notes capturing the discussion, the decisions made, and the action items that come out of it. We review the notes for accuracy and remove irrelevant content, and/or clarify where necessary. Those notes then become part of our internal notes, stored securely outside of the AI system. They play the same role typed notes always have, just with more consistency and detail, but still reviewed by humans.

What The Advisory Group does to protect your committee and your information

The Advisory Group is a fiduciary but is not part of your committee, so our notes are our own, whether taken by hand or with AI. However, notes we take could in some cases be discoverable in a lawsuit if a client is sued. As a result, we pay extra attention to maintaining appropriate notes, using the process below, and we encourage client committees to develop similar procedures and to consult an attorney to review their process for any additional guidance, and to help keep you current as regulations evolve.

• Review every set of notes. We review the AI-generated notes for accuracy and remove irrelevant or unclear content before they become part of our permanent meeting notes.
• Use only approved, private AI tools. We vet AI tools to ensure that they meet our privacy and security standards. Our notetaker is an approved private tool, which is prohibited from training Large Language Models (LLMs) and does not put meeting information into the public domain.
• Keep confidential information off public AI platforms. Our policies prohibit uploading client or company confidential information to public AI platforms.
• No storing of meeting audio. The AI system captures and processes audio conversations into text in real time. Audio is not retained as a stored file, so neither the AI tool nor we maintain an archive of meeting audio. What remains on file in the AI system temporarily are only the text notes generated from the conversation, that we tune and store outside of the AI system.
• Follow a set retention schedule. On an ERISA attorney recommendation, we retain the AI system’s original notes for nine months after the meeting date, and we have configured the system to delete them automatically once that period expires. We keep our reviewed and tuned copy of the notes, including any corrections and clarifications, indefinitely in our CRM system.

And if you’d rather we didn’t?

That’s entirely up to you. Any client or committee that would prefer we not use the AI notetaker in their meetings can simply let us know, and we’ll turn it off for those sessions. We’ll go right back to typed notes, no questions asked.

This is a developing area, and our approach may evolve as technology, the regulatory landscape, and your expectations change. We hope this gives you a clear sense of how we’re thinking about AI’s growing but carefully controlled role in our work, and the responsibility we believe comes with it. We also hope the policies we’ve developed help you shape your own policies, to improve your impact and reduce fiduciary risk.

Questions about our approach are always welcome. Feel free to reach out to our team anytime.