If you have ever closed the Airbnb inbox tab at 9 p.m. and re-opened it at 11, you already know the dread. A parking question the guest sent seventy minutes ago. A late-arrival door code that still sits unread. A refund request that you have not been awake enough to draft the reply to yet. None of it is hard. All of it is part of the dread that compiles itself between the half-marathon weekend and the wedding weekend. Roostr's 24/7 inbox auto-reply is the answer to that dread — an inbox agent that answers the routine factual messages inside the host-set policy and tone, holds refunds and schedule changes for the host's tap on the AI Suggestions hub, and keeps the inbox tab closed after bedtime. This pillar walks through what 24/7 auto-reply actually sends, where Manual stays on the credible default, and what changes on the host's 11 p.m. when the inbox tab closes for good.
What 24/7 inbox auto-reply actually means
24/7 inbox auto-reply is not a global "send everything" mode and it is not a "draft everything, queue everything" mode. It is the inbox row on the same per-surface Auto/Manual toggle the Auto/Manual pillar walks through — an Auto mode that sends the routine factual categories the host has flipped on, and a Manual mode that holds refunds, complaints, and schedule changes on the AI Suggestions hub for sign-off, regardless of how the row is set. Auto is the routine; Manual is the refund dollars and the schedule changes.
- Routine factual categories ship on Auto inside the host's policy and tone. Door code, Wi-Fi network and password, parking spot, pet fee, late-arrival instructions, checkout-time reminders, the checkout checklist — these are the categories that ship on the Auto side of the inbox row by default, every night of the week, weekend half-marathon or not. The reply lands in the host's tone, the host's policy length, and the host's tone-marker (full sentence, lowercase, emoji on/off), all set once under Settings → Automation → Inbox → Tone.
- Refunds, complaints, and schedule changes queue for sign-off, no matter how the toggle is set. A refund request, a four-star review threat, a cancellation-dispute thread, a request to swap check-in to 9 a.m. — these never send on Auto. They land on the AI Suggestions hub with the proposed reply, the reason it queued (refund dollars, complaint, schedule change), and the guest message that triggered the queue. The host opens the queue twice a day, taps approve, edits, or skips.
- Auto-send is explainable inline, every time. The reply lands in the guest's inbox the same minute the guest's message lands in the host's. The reply carries the category the agent filed it under and the policy the agent sent under. The host can flip any category from Auto-send to Review-only in a tap, and the flip is durable across the next nightly run.
- A flip survives until the host flips back. A host who flips a category off for a wedding weekend does not need to flip it back on Sunday — the host flips it back when the wedding decay has settled, and the agent stops sending that category until the host flips it on again. The flip is one tap, the flip is durable, the agent's cadence follows the host's posture.
Where Manual ships as the credible default
The inbox row defaults to Auto on the routine factual categories and Manual on refunds, complaints, and schedule changes. The default is not "Roostr sends your inbox"; the default is "Roostr sends the categories that do not touch refund dollars or guest schedule, queues the categories that do, and explains the split on the right rail of the row." A host who wants a different posture flips each category individually; a host who wants the routine runway handles the refunds.
- Every Auto-send carries the same itemized receipt a human would write. The reply lands inside the host's tone and policy; the inbox row carries the category and the policy. The host never sees a black-box "Roostr replied." The host sees the receipt and decides whether the next message of that category sends the same way.
- The host reads the queue, not the receipt. Manual-mode refunds, complaints, and schedule changes queue on the AI Suggestions hub — one row per pending recommendation, grouped by source agent. The host opens the queue twice a day and approves, edits, or skips in a tap. The receipt is on the row when the host wants it; the queue is the surface the host reads.
- Trust accumulates on the queue, not on the agent's word. A host who has reviewed thirty factual-category replies and approved thirty of them has earned the trust to add another category to Auto-send. A host who has reviewed fifty refund drafts has earned the trust to flip a category from Manual to Auto. The flip is one tap; the trust is the row-by-row review that got the host there.
- The host flips back when the math moves. A wedding weekend that lifts four Saturday inquiries gets the contact-queue category held Manual — the agent stops drafting, the queue stops teaching the host anything new for the weekend, and the host flips back on Tuesday when the lift decays. The flip is durable, the flip is reversible, the agent's cadence follows.
Built for the 1–3 listing host
Run the routine. Stay on the bigger calls.
Pricing, inbox, turnovers, and maintenance on their per-surface cadence — four Auto/Manual toggles, one workspace, the host stays in the loop on every non-routine action.
What changes on the host's 11 p.m.
The dread is not the message itself. The dread is the inbox tab re-opened at 11 p.m. between the half-marathon weekend's last inquiry and the wedding weekend's first. With the inbox row on Auto for factual categories and Manual for the rest, the inbox tab reads zero unread on a Wednesday — three routine factual messages answered in the host's tone the minute they landed, and the refund draft parked on the AI Suggestions hub for the host's morning tap.
- The parking question lands answered in the host's tone. Wednesday 9:04 p.m. The guest asks which parking spot is theirs. The reply lands in the guest's inbox at 9:05 p.m. — full sentence, the host's tone, the policy that the host set on day one. The host's tab reads 0 unread.
- The late-arrival door code lands answered before midnight. Wednesday 11:17 p.m. The guest texts about a late arrival. The reply lands in the guest's inbox at 11:18 p.m. — door-code line, arrival-instruction line, the host's checkout time. The host's tab reads 0 unread.
- The refund request queues on the AI Suggestions hub, not in the inbox. Wednesday 8:42 p.m. A guest requests a partial refund. The reply does not send on Auto. The proposed reply lands on the AI Suggestions hub with the reason inline — refund dollars, schedule change, the host-set policy the agent drafted against. The host opens the queue in the morning and approves the tap-or-skip reply in one tap.
- The inbox tab closes for the night. Wednesday 11 p.m. The host opens the inbox tab to verify. Three factual messages answered, one refund queued. The host closes the inbox tab. The dread is on the AI Suggestions hub, not on the inbox tab.
Closing — 24/7 auto-reply on the routine, sign-off on the refund
The full row-by-row matrix — every Auto-send category, every Manual sign-off queue, every flip path under Settings → Automation — lives on the Auto/Manual control deep dive and the Auto/Manual pillar. The segment landings for the two host shapes this post is shaped for are at For individual hosts and For boutique operators. The unified comparison with separately managed tools lives on /compare. Plans are on pricing.
Built for the 1–3 listing host
Run the routine. Stay on the bigger calls.
Pricing, inbox, turnovers, and maintenance on their per-surface cadence — four Auto/Manual toggles, one workspace, the host stays in the loop on every non-routine action.
If you want the 24/7 inbox running on your roster — Auto on the routine factual categories, Manual on the refund dollars — [Join the Roostr waitlist →](/waitlist).