The four Roostr automation surfaces — nightly repricing, the guest inbox, turnovers, and the review-driven maintenance loop — each produce a Manual-mode output: a draft rate change, a queued reply, a held brief, a below-threshold maintenance signal. Without a unified feed, the host would open four separate dashboard tabs to clear the manual queue. The AI Suggestions hub is the tab that replaces those four. It is a flat, newest-first list of tap-or-skip cards, one per pending recommendation, streamed from whichever surface queued it, labeled by source, and sized to a two-minute read for a single-listing host. This pillar covers what the hub actually shows, how each stream feeds it, and how to read the queue twice a day.
What the hub actually shows
The hub at /dashboard/suggestions is not a notification inbox. A notification inbox grows with every event and never shrinks until the host archives. The hub is a queue — every card that lands on it has an approve, edit-then-approve, or skip action, and the host's tap removes the card. The host who reads the hub twice a day and taps approve or skip on every card starts and ends every day with a cleared queue. The queue does not carry yesterday's rows forward unless the host explicitly skipped them.
- One card per pending recommendation. Each card shows the source agent label (Pricing, Inbox, Maintenance, Turnover), the proposed action, and the reason the agent queued it instead of executing on Auto. A Pricing card shows the comps and the event that moved the rate. An Inbox card shows the category (refund, complaint, schedule change), the proposed reply, and the guest message that triggered the queue. A Maintenance card shows the severity tier and confidence score. A Turnover card shows the brief the agent drafted and the flag that held it for sign-off.
- Approve, edit-then-approve, or skip. A tap on approve executes the proposed action — sends the reply, publishes the rate, opens the maintenance ticket, dispatches the brief. A tap on edit-then-approve opens an inline editor, lets the host adjust the draft, and executes on save. A tap on skip logs the skip and removes the card; the underlying automation resumes its cadence on the next nightly run.
- Newest first, source-labeled. The queue sorts newest-first across all four streams. The source label in the left margin lets the host scan by surface — most hosts with one to three listings clear the inbox and pricing cards first, then the maintenance and turnover cards, all in under two minutes. Hosts managing a larger portfolio can follow the scaling management company guide to build a queue-reading habit that stays ahead of volume.
How each stream feeds the hub
Every Auto/Manual surface ships with a queue behavior for its Manual-mode output. The hub is where that output lands.
- Nightly repricing. When the Pricing row is on Manual, the nightly run drafts a rate change and parks it on the hub — comps, event, and the proposed new rate on one card. The host approves, edits the final rate, or skips. The how Roostr runs the week page covers the full nightly run cadence.
- Guest inbox. Refunds, complaints, and schedule changes always queue, no matter how the inbox toggle is set. Routine factual categories queue when the inbox row is on Manual. Each queued reply shows the category, the proposed reply, and the guest message. The Auto/Manual control deep dive covers every flip path and tone setting.
- Maintenance. Below-threshold maintenance signals land on the hub as tap-or-skip cards. Above-threshold signals open directly on /dashboard/maintenance; the hub shows only the signals the per-listing threshold absorbed. The review-driven maintenance deep dive covers the threshold model and per-listing calibration, while the review-driven maintenance pillar gives the longer narrative.
- Turnovers. Held briefs — when a vetted cleaner is unconfirmed, or a listing is on Manual for turnovers — land on the hub for the host's review and dispatch.
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.
Reading the queue twice a day
The habit for a host with one to three listings: open /dashboard/suggestions in the morning and again in the evening. Clear every card — approve, edit, or skip. A morning read handles the overnight Pricing draft and any inbox queue from the previous night. An evening read handles the day's inbox queue and any Maintenance or Turnover cards that surfaced during business hours. The queue resets; the host closes the laptop with a cleared tab.
The skip action is not a disable. A skipped Pricing card does not turn off nightly repricing — the agent runs the next night at the same cadence. A skipped Inbox card does not mark the guest message as handled — the card returns on the next queue cycle until the host approves or the message thread closes. The hub tracks the host's skips so the agent can learn the host's override patterns, but the default is always to resurface an unanswered card rather than assume the skip was permanent.
The segment landings for the two main host shapes are at For individual hosts and For boutique operators. The full comparison with separately managed tools is at /compare. Plans and per-listing pricing are at /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 AI Suggestions hub running on your roster — [Join the Roostr waitlist →](/waitlist).