Reviews
Grows reviews by timing requests to after delivery and running disclosed incentives inside platform policy. Responds in your voice to 4 and 5 stars after you approve the day's batch, and routes everything under 3 stars to a human. Feeds recurring objections back into the brain.
Who it is for
The owner with 40 unanswered reviews and a review request that fires before the parcel arrives.
What it reads from the brain
Before every run. Sections are addressed as layer.section.
What it does, step by step
- Read the brain
Opens the sections above and the inputs.
- Gather evidence
Reads your public reviews and, if connected, your review app and recent orders. Every finding cites its source.
- Decide and draft
Drafts replies in your voice for 4 and 5 star reviews, times review requests to after delivery, and drafts the disclosed incentive where policy allows. Uses only claims_allowed.
- Advise stops here
You get the request schedule and the drafted replies with routing.
- Act
Sends review requests through the connected platform (Judge.me first; Judge.me, Stamped and Reviews.io can send through their APIs). Replies to 4 and 5 star reviews after you approve the day's batch. Everything under 3 stars goes to you untouched.
- Write back
Recurring objections and vocabulary from reviews are written to the brain; what did not work is appended.
Tools it uses
storefront.reviewsPublic reviews at level 1judgeme.reviews.list, .reply, .review_request.sendAct: read, reply (approval gated) and request (approval gated) on Judge.mestamped.reviews.list, .review_request.sendAct: the same on Stampedreviewsio.reviews.list, .review_request.sendAct: the same on Reviews.ioshopify.orders.listDelivery timing for requests, one per order
The three modes for this agent
Read tools only. You get the drafted replies with star rating and routing, and the request schedule timed to delivery.
Requests sent through your review app after approval. Replies to 4 and 5 stars sent after you approve the day's batch. Under 3 stars: routed to you, never answered by the agent.
Every day at 09:00. Same gates. Requests go out after every delivery; replies are drafted every morning for your tap.
Guardrails and approvals
Enforced in code. Items with a filled circle wait for your tap in Telegram.
- Never gates negative reviews. That is illegal.
- Never replies below 3 stars without a human.
- Never offers a refund or discount in public.
- One review request per order.
- Incentives disclosed, inside platform policy.
- Google review text is never stored.
- Loox and Okendo are read only; no requests can be sent through them.
- Every claim in every reply is in claims_allowed or is removed and listed as needs your confirmation.
What it writes back to the brain
Dated and sourced. Learnings and performance sections are append only.
Example output
Example output. Structure only. Counts are placeholders.
Reviews, your store, today
Ran at data level 3 (Judge.me and Shopify connected).
- New reviews since yesterday: [n]. 4 and 5 stars: [n], replies drafted below. Under 3 stars: [n], routed to you untouched, in escalations.
- Review requests timed to delivery today: [n], one per order.
- Recurring objection added to the brain: [the phrase customers used], seen [n] times this month.
Waiting for your approval on today's batch of [n] replies.
replies.md (one entry)
5 stars, [date], hero product 1. Customer wrote: [their words]. Drafted reply, your voice: Thank you for [the specific thing they praised, in their words]. [One line inside claims_allowed about how to get the most from it.] Routing: send after approval.
2 stars, [date], hero product 2. Routed to you. No draft. No public reply from the agent.
Also delivered: requests.csv (order_id, delivered_at, request_scheduled_at, platform, status).