Decide once. Let the rules do the rest.
Priority-ordered rules decide what publishes on its own and what waits for a person. Approval is a switch you control, not a step you're stuck with — and nothing fails silently.
Replying by hand doesn't scale — and neither does replying blindly
The choice is usually framed as all-manual or all-automatic. Both are wrong, and both break in the same place: the reviews that actually need a person.
Every reply is a manual decision
Reading each review, deciding whether it is routine, and typing a response is work that scales exactly with review volume — which is the wrong thing to scale with.
One blanket policy for everything
Turning automation fully on posts replies you would rather have read first. Turning it fully off means the five-star reviews wait behind the hard ones.
Failures happen quietly
A dropped Google connection or a rejected publish leaves reviews silently unanswered, and nobody finds out until someone opens the profile.
From a new review to a published reply
Five steps run every time a review arrives. You configure steps two and three; the rest happens whether you are watching or not.
Review syncs in
A new review lands on your Google Business Profile and arrives in the inbox.
Rules evaluate
Your conditions run in priority order. The first rule that matches wins.
Outcome applied
Auto-reply, require approval, hold, or skip AI for this review entirely.
Guards check the draft
Validators catch risky language before anything is allowed to publish.
Publish or alert
The reply posts to Google. If it fails, it retries and your team is told.
Conditions in priority order, first match wins
A rule pairs a condition — a star rating, a keyword in the review text — with one outcome. Rules are evaluated top to bottom and the first one that matches decides what happens, so a review always gets exactly one outcome and you can read the stack like a sentence.
The four outcomes
Reordering the list is how you resolve an overlap. Nothing runs twice, and no review falls through — the last rule is the catch-all.
- Rating is 5 starsAuto-reply
- Review mentions refund or billingRequire approval
- Rating is 1 or 2 starsHold for a human
- Review has no comment textSkip AI
- Anything elseDraft and queue
Approval is optional, and that is the point
Every location carries its own auto-reply mode. Full-auto and human-in-the-loop are both first-class settings — neither is the "real" one you are meant to graduate to.
Off
Nothing is drafted and nothing is published for this location. The inbox still syncs, so replies stay a manual job by choice rather than by accident.
Draft only
Every reply is written for you and parked in the approval queue. A person reads, edits if they want to, and decides what goes out.
Full auto
Drafts publish to Google on their own, with no queue in between. This is a supported end state, not a shortcut you are talked out of.
Rules override the default in either direction
The mode is the location's default, not a ceiling. A rule can hold a review at a branch that otherwise runs fully automatic, and a rule can publish automatically at a branch that otherwise queues everything. That is how one estate ends up with routine praise answered in minutes while anything mentioning a refund still waits for a person.
For everything you did want to read first
When a rule or a location mode holds a reply, it goes here — with the review it answers, ready to edit or send.
One queue for everything held
Whatever a rule held — or whatever a draft-only location produced — lands in a single list with the review it answers sitting next to it.
Edit in place
Change the wording right in the queue and publish the edited version. There is no separate compose screen and no copy-paste round trip.
Bulk approve
Select a run of drafts you are happy with and send them all at once, instead of clicking through them one review at a time.
What stops a bad reply, and what catches a failed one
Two things run underneath every mode, including full-auto — because the case for letting replies publish on their own only holds if something is checking them.
Guard validators
Every draft is checked before it is allowed to publish. Risky language — admissions, promises, anything that reads badly in public — is flagged rather than posted.
Automatic retry
A publish that fails is not dropped. It is retried on its own, and the reply keeps its place until it is genuinely on Google.
Routed per trigger, to wherever your team already is
Each trigger is routed on its own, so the channel that gets a 1-star review does not have to be the channel that gets a connection error at 3am.
- A 1-star review arrives
- A negative review arrives
- A review is still unreplied
- A reply fails to publish
- A Google connection breaks
Slack
Post into the channel the team already watches.
Discord
Same routing, for teams that live in Discord instead.
For people who are not in a chat tool all day.
Signed webhook
A signed POST to your own endpoint, so you can route it anywhere.
What a configured month can look like
A sample workspace running the example rule set above.
- Published automatically
- 68%
- Held for approval
- 32%
- Median time to reply
- 6 min
- Retried and recovered
- 4 replies
Illustrative figures from a sample workspace, shown to explain the shape of the split between automatic and held replies. They are not customer results and not a projection — your own split depends entirely on the rules you write.
Before you switch anything on
The four things people ask before they let a reply publish itself.
Do I have to approve every reply?
No. Approval is optional and always has been. A location can run fully automatic, draft-only, or off, and you can change your mind at any point. Plenty of teams start with everything queued, then move the straightforward five-star reviews to automatic once they trust the wording.
What happens when two rules could both match?
The higher one, always. One review gets exactly one outcome, so an overlap is never ambiguous — it just means the rule you want is sitting in the wrong place, and you drag it up.
Can a rule override a location's setting?
Yes, in both directions. The per-location mode is a default rather than a ceiling, so a matching rule decides the outcome whichever way it points.
What happens if publishing to Google fails?
The reply is retried automatically rather than quietly dropped, and a publish failure is one of the triggers you can route to Slack, Discord, email, or a signed webhook — so a broken connection surfaces as a message rather than as a gap you notice weeks later.
Where to next
The parts of the platform that sit either side of the rules engine.
Write the rules once. Stop typing the replies.
Start with everything queued and move things to automatic as you trust them — or go hands-off on day one. Both are supported, and you can change your mind per location at any time.