Last updated:
Quick answer
Write a one-page voice guide: tone, length, sign-off, escalation list, banned phrases. Make it the system prompt for AI drafts so everyone who replies sounds like one business.
At a glance
All outcome figures are estimates from cited sources, not independent verification by us.
- What it is allowed to do
- Drafts for you
Writes drafts (emails, posts, plans, replies). You review and send. Nothing goes out without you.
- Time saved (estimate)
- ~3 hours per month plus fewer brand fires
- Difficulty
- Advanced
- Tools needed
- Claude, ChatGPT
- Typical cost
- Chat tools on any tier. Claude Projects or ChatGPT custom instructions keep the guide persistent.
Reviewed Oct 19, 2026
Basis (estimate): Estimate basis: assumes inconsistent replies cost ~1 hour/week in rewrites and manager cleanup vs a shared standard with AI drafts.
What does this setup do?
Create a shared review-response standard for teams: who replies, how they sound, what escalates, and how AI drafts stay on-brand.
Who is it for?
Marketing managers, ops leads, and multi-location owners where two or more people answer reviews and the tone swings wildly.
How do you set it up?
Tools
- Claude or ChatGPT (a Project or shared doc keeps the guide as the standing instructions)
- 20 recent replies from your team, good and bad
- Your brand facts: voice, forbidden phrases, escalation contacts, legal or compliance limits
Steps
- Collect 20 recent team replies. Mark the 5 best and 5 worst, and note why in one line each.
- Define the voice in plain words: formal or casual, first names or not, emoji or not, typical length.
- Write the forbidden list: phrases nobody uses ("valued customer," "we regret any inconvenience"), plus anything legal or compliance bans.
- Set the escalation list: health claims, legal threats, employee names, press mentions. Those always go to a named person.
- Turn the guide into the system prompt for AI drafts: voice, length, sign-off, escalation triggers, forbidden phrases.
- Name one approver for tricky replies. AI drafts, the owner or manager approves anything flagged.
Example output
Illustrative. Your results will differ.
- 5 new reviews found: 4 positive, 1 needs a reply
- Reply draft ready for the unhappy one
- You approve every reply before it posts.
Where to paste this prompt
Claude: Claude → Project knowledge & custom instructions
Copy-paste prompt
Adapt bracketed fields to your situation.
Write a one-page review response playbook for my team. Concrete and enforceable, not motivational fluff.
Business: [name, locations]
Team: [who replies: e.g. 2 store managers, 1 marketing lead]
Voice: [e.g. warm but professional, first names, no emoji]
Regulated or sensitive topics: [e.g. HIPAA, financial advice, alcohol service / none]
Here are 5 of our best replies and why they worked:
[paste]
Here are 5 of our worst and what went wrong:
[paste]
Return the playbook:
1. Voice rules (5 bullets max).
2. Length rules: positive reviews, negative reviews, mixed reviews.
3. Sign-off format and who signs what.
4. Forbidden phrases list.
5. Escalation list: what always goes to [named approver] before posting.
6. The AI draft instructions block: paste-ready system prompt encoding all of the above.Newsletter
Get the free starter pack
Free starter pack with five beginner copy-paste prompts after you confirm, plus three sourced setups in your inbox each week. Unsubscribe anytime.
What usually goes wrong?
Where agents tend to stall on this job. More patterns in Where AI gets stuck.
- A guide nobody reads is decoration. Walk the team through it once with real examples.
- Over-automation: full auto-reply without human review eventually posts something tone-deaf. Keep a human on negatives.
- Regulated industries (health, finance, legal) need compliance review of the forbidden list, not just marketing.
- Templates pasted verbatim across locations get spotted. The guide should demand personalization, not just allow it.
Which chat tools work with this prompt?
Paste the prompt into any of these assistants (free or paid tiers).
- Claude
- ChatGPT
Common questions
- How long should the playbook be?
- One page. If it needs a table of contents, nobody will follow it. Voice rules, length rules, forbidden phrases, escalation list, done.
- Can AI fully automate our review replies?
- For simple positives, semi-automation with human spot-checks works. For negatives and anything flagged, keep a human in the loop. One bad auto-reply costs more than the time saved.
- Who should own the playbook?
- One named person, usually marketing or ops. They refresh it quarterly and approve escalated replies.
What tips make it smoother?
- The playbook is also the AI’s instructions. Write it once, use it for both humans and drafts.
- Review the guide quarterly against real replies. Voice drifts; the doc should catch it.
- Keep approved openers and sign-offs in the guide so locations sound like one brand.
- Track who handled escalations and how they resolved. That log becomes training material.
Where do the claims come from?
Summaries use public sources linked below, not independent verification. How setups are picked.
- Read ShoutAboutUs’s write-up (www.shoutaboutus.com)
ShoutAboutUs: What belongs in a reply style guide: voice, length, sign-off, contact route, escalation, forbidden content.
- Read GBP Central’s write-up (gbpcentral.com)
GBP Central: Tone consistency principles: warm but professional, owner’s voice, proportional, no visible templates.
Related setups
- Morning inbox briefing and multi-inbox triage
Start the day with a short “what needs you” note, and periodically sort multiple inboxes into reply vs FYI vs junk, with drafts for your approval, not auto-send.
- Prep for meetings from the right email thread
Before a meeting, find the relevant email thread, summarize status and open items, and cross-check mail against your calendar for conflicts.
- Track recurring software spend and idle charges
Scan renewal emails and receipts for SaaS and AI tools, flag overlaps, and draft refund or cancellation messages for charges you are not using.