Every preference, allergy, anniversary and pet name is lifted out of real conversations the moment it is said, on any channel, in any language, and written to a structured guest profile every agent can read.
The moment a guest writes, calls or arrives, the relevant memories are pulled from their profile and folded into the reply before it is composed. Same store, every surface, no template.
Being asked the same question a third time is the moment a loyal guest becomes a transaction. Memory is the difference between a hotel that recognises someone and a system that processes them.
Captured without anyone typing it, filed into a slot the agents already know how to read, traced back to the exact line it came from, and available anywhere the guest is talked to or talked about.
Every sentence a guest writes is sorted into a first-class category the agents, segments and journeys already know how to read. Watch a live line land in its slot.
Nothing in a guest profile is a black box. Every memory carries the channel, the timestamp and the original sentence, and your team can correct it, pin it or forget it in one click.
"Returning pet-parents with an anniversary in the next month" is a memory query. Segments treat memories as first-class filter fields, and every journey step re-reads them at send time, never a stale template.
See what is being remembered, which categories are thin, and how often a memory actually changed a reply, a journey or an offer. Coverage you can act on, not a vanity counter.
A note is a paragraph nobody reads. A memory is a typed record: a category, a structured value, a confidence score and the exact message it came from. Because it is typed, agents can act on it, segments can filter on it and journeys can personalise with it, none of which a free-text note allows.
Every inbound conversation: WhatsApp, voice call transcripts, email threads, web chat, Instagram and Facebook DMs, plus PMS and staff notes. Nobody on your team has to type anything, though staff can add or correct a memory by hand at any time.
Every memory carries a confidence score and its original source. Low-confidence extractions are held for review rather than used. Anything already saved can be corrected inline or forgotten in one click, and forgetting removes it from every agent, segment and journey immediately.
A stated memory is something the guest actually said. An inferred one is a pattern we observed, such as the same companion across four stays. Inferred memories are always labelled as such, and anything the guest states directly overrides them.
No. Memories live in your tenant only and are never used to train a shared model. Deletion requests are honoured end to end, across the memory store and everything that reads from it.
Yes. There are 38 first-class categories out of the box, and you can add custom ones per property. Custom categories behave exactly like the built-in ones: queryable from segments, readable by agents, usable in journeys.
They are a deliberate pair. The Hotel Brain holds what the hotel knows, its facts, prices and policies. Guest Memories holds what the guest told you. Together, every reply is both factually correct and personally relevant.
Thirty-minute demo. We'll run your own conversations through the memory extractor in front of you, show you what it lifts out, and open any memory back to the exact line it came from.