Messaging
AI reputation manager that answers Google reviews safely and reports every week
Reputation Manager reads your Google Business Profile reviews and, if connected, Instagram and Facebook comments, drafts a reply to each in your brand's tone, posts only positive, non-risky 4–5 star replies when you allow it, routes 1–3 star and risky reviews to your approval and Inbox, and writes a reputation summary for the period.
- As a workflow node
agent.reputationManager- Default model
- Claude Sonnet 5
- Economy model
- Claude Haiku 4.5
- Tools
- 6
- Report
- REPUTATION_REPORT
What the agent does
Replies are short: they thank the reviewer by display name, echo one detail from the review and use no links, hashtags or sales talk. When facts such as opening hours, booking or returns are needed, the agent searches your knowledge base first and writes only what it finds. Reviews that mention health or safety, legal threats, discrimination, personal data or an escalation keyword are risky and are never answered automatically, whatever the stars.
The platform never posts automatic replies to 1–3 star reviews: the agent writes a calm, apologetic draft, hands the review to you with a severity and the draft waits on the Approvals page. The agent never confirms who a customer or patient is, never mentions order, visit or treatment details, never offers discounts or gifts for reviews and never asks anyone to change a review.
The period summary covers 3–8 themes customers praise or criticise, the average rating compared with the previous period, the star distribution, reply stats and action items with owners; it is saved on the Reports page as a Reputation summary and announced to you. For hourly comment replies, Community Manager is the better fit.
Inputs / Outputs
Inputs
- Platforms
- Google Business Profile reviews (default) and, if connected, Instagram and Facebook comments.
- Auto-post 4–5★ replies
- When off (default), every reply is a draft on the Approvals page. When on, only positive, non-risky 4–5★ replies are posted right away; 1–3★ and risky reviews always wait for approval.
- Summary period
- Last 7 days (weekly) or last 30 days (monthly). The agent also reads the period before it for the rating trend.
- Escalation keywords
- Words that make a review never auto-answered and handed to you; add terms such as
refund,complaint,food poisoningorallergy. - Reply tone
- Warm, professional, friendly or formal. The brand voice in your project settings still applies.
Outputs
- replies
- A reply per review: platform, author, excerpt, reply text, stars, sentiment, risk flag and whether it was posted; unposted ones wait on the Approvals page.
- escalations
- Reviews handed to you: platform, stars, reason and severity.
- themes
- Topics customers praise or criticise: sentiment, how often they come up and a short anonymous example.
- ratingTrend
- The period's average rating against the previous period, review count, star distribution and direction (up, down, flat).
- replyStats
- How many reviews were processed, posted, left as drafts, handed to you and skipped.
- actionItems
- Prioritised actions: what to do, why and who owns it.
Saves a report to the Reports page on every run: REPUTATION_REPORT.
Tools
- Read comments
get_comments - Reply to comments
reply_comment - Sentiment analysis
classify_sentiment - Hand off to a human
escalate_to_human - Knowledge base search
knowledge_search - Notify the owner
notify_owner
Integrations it uses
Example tasks
- “Answer this week's Google reviews; thank the 5★ ones with a detail they mentioned and send me the 1–3★ ones with a draft.”
- “Summarise the last 30 days of reviews: what do people praise, what do they complain about, and what should we fix first?”
- “Reply to new reviews in a formal tone; never mention prices or offer discounts, and invite unhappy guests to e-mail us.”
Tips for good results
- Run it on a weekly schedule; the summary is ready on the Reports page every Monday.
- Add opening hours, booking and return policies to the knowledge base; the agent never writes what is not there.
- Keep auto-posting off for the first weeks and turn it on once you trust the drafts.
- If you run a clinic, add “never confirm someone is a patient, never name a treatment” to Additional instructions; the agent already follows it, and it sets the rule for your team too.
Model advice
Default model
Claude Sonnet 5
Per 1M tokens: $2 input · $10 output
Economy model
Claude Haiku 4.5
Per 1M tokens: $1 input · $5 output
The default balanced model is recommended for getting the tone right on negative or sarcastic reviews and for grouping themes meaningfully. If most reviews are short thank-yous, the economy model is enough; sentiment classification always runs as a separate, low-cost economy-model call.
Templates that use this agent
Frequently asked questions
Does it auto-reply to 1–3 star reviews?
No. For these reviews it only writes a draft and hands the review to you; the reply goes out when you approve it on the Approvals page.
Does it offer discounts or gifts for positive reviews?
No. Incentives for reviews break Google's policies and advertising rules; the agent never does it in any reply.
Does it delete or hide reviews?
No. The agent only writes replies and routes reviews to you; it flags suspected fake reviews, and reporting them is your call.
Get started with OrqLabs
Build your first workflow today
Start on the free plan: connect your accounts, pick a template and stay in control with approval steps. No credit card required.