The AI Agent starts every conversation already knowing something about the guest it is talking to.
Before it composes a reply, it receives a short guest context summary — a few sentences describing that guest in patterns and preferences, built from their master profile. The result is a reply that reads more like a front-desk colleague who recognises a returning guest than a generic response.
Previously, every conversation started from a blank slate. A guest of ten years, a family with a nut allergy, and a first-time visitor all received the same tone. Guest context gives the AI Agent the background your property has spent years building — without ever handing it the guest's personal data.
Nothing to configure. If your AI Agent is active, guest context is active.
What it draws on
The summary is built from the guest's master profile — their clustered identity across every sub-profile and reservation you hold — together with:
Stay history and loyalty tenure
Segments and guest tags
Travel patterns
CRM notes
Consents
Form and submission activity
It comes out as two to six sentences along the lines of "returning guest, several stays, travels as a family, has a nut allergy" — patterns, not records.
Colour, not facts
This is the most important thing to understand about the feature.
Guest context shapes tone and background only. It never supplies a verifiable detail. Anything a guest could hold you to — check-in status, dates, amounts, room numbers, booking references — always comes from a direct lookup, never from the summary.
Numbers are deliberately coarsened for the same reason. The summary says "has stayed several times", never "has stayed four times", so nothing in it can be quoted back as authoritative.
For reservation details specifically, see Answering reservation questions: how guest verification works.
What this changes in practice
Returning guests are recognised. The AI Agent adjusts its tone to match their history with you rather than treating a loyal guest as a stranger.
Care information surfaces. Allergies and accessibility needs are handled with the right sensitivity, without the guest having to explain again.
Travel patterns inform the reply. Business or leisure, family travel, typical stay length — replies land closer to what the guest actually needs.
CRM notes are reflected. Preferences, special requests, and prior complaints shape the reply without an operator having to repeat them.
How guest privacy is protected
The summary never contains personal data. Every generated summary is put through an automatic check before the AI Agent sees it, looking for names, email addresses, phone numbers, and booking references.
If any of those survive into the summary, the entire summary is discarded and the AI Agent replies without it. There is no partial redaction — it is all or nothing, so nothing slips through.
The summary is also never shown to the guest. It only shapes how the AI Agent writes.
Which guests get it
Only guests who have a master profile in the search index. Brand-new guests, and guests whose sub-profiles have not been clustered into a master profile, get no context.
This is not an error state. If there is no cluster, no data, or a rejected summary, the guest still gets a normal, correct answer — just a less tailored one.
What you will notice
Replies vary between guests. The same question from two different guests can come back phrased differently. That is the feature working, not your tone setting drifting.
A brief typing indicator. Building the summary adds a step, covered by a typing indicator on Web Messenger and WhatsApp.
Nothing new in the inbox. The summary is not shown on the thread. Your team sees the AI Agent's reply, not the context behind it.
Seeing it for yourself
Admins can preview the behaviour in the Preview panel. Go to Inbox Settings > Bots, open your AI Agent, and use the Testing tab and Preview panel as described in Testing Your AI Agent with the Preview Panel.
The testing panel has a guest selector, which composes a real guest's context through the same path the live AI Agent uses. Picking a returning guest and a first-time guest and asking the same question is the clearest way to see the difference.
This is currently the only place the summary's effect can be observed directly.
Current limitations
Up to 24 hours stale. Summaries are cached per conversation for 24 hours to keep replies fast. A CRM note added minutes ago may not be reflected until the cache expires. This is a deliberate trade-off — there are around 19 million master profiles behind this.
Guests must be clustered. No master profile means no context.
No exact figures, ever. Stay counts, dates, spend, and age are intentionally coarsened.
Not surfaced in the inbox yet. Operators cannot see the summary on the thread.
All-or-nothing on privacy. A single detected identifier drops the whole summary for that reply.
Need Help?
Please contact us through the 'Talk to Us' option on the left menu in the platform, or through the Bookboost Support email at support@bookboost.io if you have questions or need additional support.