What is meeting memory?

Meeting memory is the ability of an AI support agent to recall what a returning customer discussed in past meetings with your team: dates, interests, commitments and summaries, drawn from meeting notes rather than from the chat history. It is different from session memory, which is the model keeping earlier turns of the current chat in its context window, and from a CRM lookup, which returns fields rather than what was said.

It matters because customers hate repeating themselves, and a demo call, a pricing worry or a promised follow-up are the things they most expect a company to remember. An agent that can say "on 18 September you discussed the Growth plan with Dana" turns a cold chat into a continuation. The idea is young and the name is not yet standard, so expect vendors to call the same feature something else or to fold it into "customer context".

Where it runs out, and should: memory must be gated by identity, or anyone typing a customer's email reads their meetings. It should carry summaries and commitments, not transcripts, recordings or details about other attendees. It should be off until a business turns it on, and it should never invent quotes, because summaries are approximate. A memory feature without those limits is a privacy incident waiting for the right visitor.