Three reply behaviors that make the concierge feel like a thoughtful concierge instead of a search field: a one-sentence affirmation before results, transparent dates when the guest is vague, and a row of click-to-apply alternative dates when the picked window is loose.
What it does #
When a guest searches in the chat, the concierge can wrap the result list in three small but noticeable touches.
A grounded affirmation, not “Here are your results.” Before the cards render, the concierge can lead with a one-sentence acknowledgement of what the guest asked for, tied to a true fact about your destination. For example, on an anniversary trip search to Cocoa Beach in March, the concierge may open with “An anniversary trip to Cocoa Beach in March, shoulder weeks tend to feel calmer than peak.” Where it has a relevant fact in your Concierge Local Knowledge, it uses that; otherwise it falls back on built-in booking dynamics (inventory size, lead time, weekday versus weekend). It stays silent on a search that returned nothing so the guest is never affirmed for an empty result.
Transparency on fuzzy dates. When a guest gives a loose date (“in July”, “mid-week”, “this summer”), the reply always states the stay length and the exact dates it picked, for example “a 3-night mid-week stay, July 14-17.” The guest never has to guess what the concierge actually searched.
Pre-validated alternatives. When that fuzzy-date search returns results, the concierge can also offer a small set of alternative date chips below the affirmation: nearby start dates, and shorter or longer stays. Each chip is checked for availability before it is shown and is labeled with how many homes are available. The strongest option leads. Chips are click-to-apply, so the guest can switch windows in one tap.
What the guest sees, in order:
- A one-line acknowledgement of their search (when on).
- The chosen window, stated plainly when their dates were loose.
- A row of click-to-apply alternative-date chips (when the window was loose and the search returned results).
- The matching property cards.
How to turn it on #
1. Open the AI Concierge admin #
Go to <strong>HomeRunner -> AI Concierge -> Chatbot</strong>.
2. Decide whether to affirm searches at all #
Toggle <strong>Contextual response prefix</strong>. On by default. Off keeps replies terse and transactional.
3. Pick how often the affirmation fires #
Set <strong>Affirmation – when to fire</strong> to one of:
- <strong>First search only</strong> – affirm the opening search, stay terse after.
- <strong>Substantive searches</strong> (recommended) – always affirm the first search, then only the effortful ones (a longer message or three or more filters).
- <strong>Every search</strong> – affirm every turn.
4. Pick where the reason comes from #
Set <strong>Affirmation – reason source</strong> to one of:
- <strong>Prefer Local Knowledge, fall back to generic</strong> (recommended).
- <strong>Local Knowledge only</strong> – stay silent when there is no grounded fact for this search.
- <strong>Generic booking dynamics</strong> – skip Local Knowledge entirely.
For the affirmations to feel specific to your destination, fill in your Concierge Local Knowledge on the Concierge tab (town, seasons, trip types). The richer the Local Knowledge, the more often “Prefer Local” wins.
5. Decide whether to offer alternative-date chips #
Toggle <strong>Fuzzy-date alternatives</strong>. On by default. Turn it off to keep loose-date searches to a single picked window.
Settings #
<!– table –>
| Setting | What it controls |
|---|---|
| <strong>Contextual response prefix</strong> | Master switch for the one-sentence affirmation before results. Default on. Always skipped on zero-result searches. |
| <strong>Affirmation – when to fire</strong> | How often the affirmation appears: <strong>First search only</strong>, <strong>Substantive searches</strong> (default), or <strong>Every search</strong>. “Substantive” always affirms the first search of a session, then only effortful later turns, so quick follow-ups stay clean. |
| <strong>Affirmation – reason source</strong> | Where the affirming reason comes from. <strong>Prefer Local Knowledge, fall back to generic</strong> (default) uses a true sentence grounded in your Local Knowledge when one applies, otherwise a generic booking-dynamics line. <strong>Local Knowledge only</strong> stays silent without a grounded fact. <strong>Generic</strong> uses booking-dynamics phrasing only. |
| <strong>Fuzzy-date alternatives</strong> | Pre-validated alternative-date chips shown below a successful loose-date search. Default on. Each chip is availability-checked before display and labeled with how many homes are available; the strongest option leads. Off keeps successful fuzzy-date searches to a single picked window. |
Tips #
- <strong>Fill in Concierge Local Knowledge first.</strong> The affirmation feels concierge-grade when there is a real fact to ground it in. With Local Knowledge empty, “Prefer Local” simply falls back to generic phrasing.
- <strong>”Substantive” is the right starting point.</strong> “Every search” can feel chatty on quick follow-up turns; “First only” misses obvious affirmation opportunities later in a session.
- <strong>The alternative chips never offer a worse window than the current one.</strong> If no better window is available, the row simply does not appear. Operators do not need to filter or curate the chips.
- The fuzzy-date chips show an approximate count (“a few available”, “10+ available”) rather than an exact number, because the grid clusters same-building units; ranking still uses the exact internal count.
- See <a href=”/docs/ai-concierge-property-chat/”>Property Chat</a> for how Local Knowledge powers single-property answers, and the <a href=”/docs/homerunner-ai-concierge-overview/”>AI Concierge overview</a> for where Smart Replies sits in the larger picture.
</content>
</invoke>