draft-review-replies-in-airtable: native input mapping
Turn new reviews into response drafts

Purpose and compatibility
A blank file is an input scaffold, not a native export or completed mapping. Preserve the exact column names/order and input names. Populate only from the documented native acquisition or explicit manual evidence. Stop if a required input, permission, privacy condition or coverage requirement is missing. Do not replace an old prompt until the named inputs and field-level mapping are reviewed. These files do not authorize login, setup, imports, account changes, message sending, publication, test submissions or spending.
Contract status: Requires distinct native-input templates or an explicit reviewed mapping; existing kit is not drop-in compatible
Compatible without mapping: false

Required inputs
SanitizedReview -> native-sanitizedreview-template.txt
Required content: Human-approved non-sensitive review excerpt, editorial review alias and genuine review source reference; sensitive/unsanitized input is blocked.
Complete each required field from approved source evidence.

ApprovedReplyPolicy -> native-approvedreplypolicy-template.txt
Required content: Approved tone, prohibited disclosures, escalation triggers and named reviewer role.
Complete each required field from approved source evidence.

Scope -> native-scope-template.txt
Required content: Full approved scope, source account aliases, date filters, date basis/timezone, as-of capture time, definitions, source-to-alias map provenance, matching rules, inspected/excluded records, permission/privacy limits, missing-data reasons, coverage and responsible reviewer. Include recipe-specific policies, versions and tests. Scope must be inside the AI workbook or pasted packet, not merely a separate unprovided file.
Complete each required field from approved source evidence.

Old files and reviewed mapping
input-preparation.txt
Original first line: Before opening the official Make template, identify the test location placeholder, test Airtable table, reviewer role and approved response policy. Review the available connector fields for review text, source reference, response draft and pending-review status. Keep the scenario inactive. Verify account ownership, connection permissions, plan limits and OpenAI usage costs before a separately approved connection. If the staff approval gate cannot be represented, use a manual review queue and stop automatic publication.

response-policy-template.txt
Original first line: Policy owner: [required]

response-policy-template.txt -> ApprovedReplyPolicy: carry the approved tone, prohibited disclosures, escalation triggers and named reviewer role. input-preparation.txt -> Scope: verify real account/module fields, target-table permissions, privacy review, costs and human-review gate.
SanitizedReview requires a genuine review source reference, editorial alias and separately human-approved non-sensitive excerpt. Raw public legal-review text can itself be sensitive. The old instructions and synthetic examples provide no genuine review or authorization to send text to AI. Keep the Make scenario inactive.

How to perform the handoff
1. Keep original exports and raw evidence unchanged in restricted firm storage. Check actual source fields, dates, coverage, privacy and access before mapping.
2. Prepare the distinct native inputs above. Record every mapped field, derived interpretation, missing value and observed source reference in Scope. Identically named columns still require source/evidence validation.
3. Create a separate minimized analysis workbook or text packet. Name tables/sections exactly as listed. A CSV file does not itself create a named Excel table.
4. Include the full completed Scope in the same AI-visible workbook or pasted packet. Verify exclusions, denominator, matching rules and unknowns. Do not hide raw client data on another worksheet.
5. Obtain the responsible reviewer’s approval of the completed mapping and evidence coverage before using the native prompt. Keep all blocked items visible. Any later external action is a separately authorized human step.

Acquisition and privacy requirements
Use approved minimized administrative metadata and aliases only. Aliases alone do not guarantee anonymity. Keep untouched exports/screenshots, identifying source maps, full documents and sensitive narratives in restricted firm storage. A hidden worksheet is not a privacy boundary. Do not insert credentials, payment/tax identifiers, raw client documents or unsanitized review text. Unknown is not zero, false, completed or absent.
Existing authorized access to the correct source account and records; no new login, credentials, installation or permission grant is included.
Responsible lawyer/operations reviewer approves scope, definitions and any later action.
AI optional; use only approved minimized data and retain originals separately.
Official template is a reusable starting point, not a tested blueprint. Safe legal use requires a sanitization/review gate absent from the claimed default description.
The live account, entitlement, exact headers, records and field coverage have not been observed.

Recipe acquisition / manual mapping instructions
Identify the exact template and safe input boundary
Location: Official Make template 15168; GBP listing > Read reviews; Google Maps review > More > Share review for source URL
Read template 15168: Google Business Profile reviews > OpenAI response generation > Airtable records. For a safe manual pilot, capture an actual public review link and only lawyer-approved nonsensitive wording. Public reviews can themselves reveal sensitive facts; do not route raw review text automatically to AI.
Verification: Template existence proves neither installed module versions nor tested field mappings. It was not imported or run.
https://www.make.com/templates/15168-monitor-google-my-business-reviews-and-create-records-in-airtable-with-chatgpt-responses
https://support.google.com/business/answer/3474050?hl=en


Do not proceed when
- Missing approved reply policy, unsanitized sensitive review, unverified connections/module fields or target-table permissions.

Scope template requirements
Full approved scope, source account aliases, date filters, date basis/timezone, as-of capture time, definitions, source-to-alias map provenance, matching rules, inspected/excluded records, permission/privacy limits, missing-data reasons, coverage and responsible reviewer. Include recipe-specific policies, versions and tests. Scope must be inside the AI workbook or pasted packet, not merely a separate unprovided file.

Old examples and validation limits
Legacy synthetic examples and negative cases belong to their old prompts. They do not automatically test the native route and must not be fed into a new prompt without an explicit reviewed mapping and separately approved test design. These new files contain no synthetic values, source exports, account results or live-test evidence.
Documentation-checked component path; proposed workflow, not tested
Public official documentation checked 2026-10-15; no login, export, live query, import, installation, configuration, message or client workflow was executed.
Local validation covers only exact scaffold schemas, placeholder presence, blank CSV rows and required warnings. It does not establish live account coverage, native export validity, successful integration or completed workflow tests.
