build a structured customer support content library.
You bring your real customer questions and business context; the AI helps you organize them into categories, find gaps in your coverage, write clear answers in your brand voice, and plan how to deploy everything across your support channels.
Customer Support Content Planner
Tags: customer support, FAQ, chatbot planning, content strategy, knowledge base, self-service, automation planning, beginner, customer service
TL;DR
What It Does
Walks you through building a structured customer support content library. You bring your real customer questions and business context; the AI helps you organize them into categories, find gaps in your coverage, write clear answers in your brand voice, and plan how to deploy everything across your support channels.
How It Works
Four interactive phases, each with a pause for your review: (1) Question Inventory — organize what customers ask, (2) Gap Analysis — find questions you missed, (3) Answer Writing — draft clear responses in your brand voice, one category at a time, (4) Organization and Deployment Guide — structure your content library and plan channel deployment.
What You Need
Your business type, what you sell or offer, and real customer questions (the more specific, the better). Optionally: existing FAQ or help content, a description of your brand voice, and which support channels you use.
Typical Time
30-60 minutes. This is an interactive planning session — the pauses between phases are where your edits and additions create the most value.
How To Start
Collect your real customer questions before starting. Actual support emails, chat logs, or ticket subjects produce significantly better results than summaries.
STEP 1Gather Your Support Knowledge
Before running this recipe, collect information about what your customers actually ask. Paste real support emails, chat logs, or ticket subjects. List your most common questions. Note which questions have simple answers versus which need back-and-forth. If you have an existing FAQ page or help docs, copy that text too. Even 10-15 real customer questions will produce significantly better results than a vague description.
Available parameters
-
business_type
· string · required
Your business type and industry. Be specific: “Online store selling handmade jewelry to US consumers” works better than “e-commerce.” -
products_services
· string · required
What you sell or offer. Include details that affect support: pricing tiers, subscription models, physical vs digital delivery, warranty or guarantee policies. -
known_questions
· string · required
Customer questions you already know about. Real examples (pasted emails, chat logs, ticket subjects) produce the best results. -
existing_content
· string · optional · default “none”
Your current FAQ page text, help docs, or canned responses. Paste the full text so the AI can identify gaps and suggest improvements, or “none” if starting fresh. -
brand_voice
· string · optional · default “friendly and professional”
Your communication style. An example sentence works better than a label: “We write like we’re texting a friend who asked a smart question” beats “casual.” -
support_channels
· string · optional · default “email”
Where you deliver support or plan to deploy this content: email, live chat, FAQ page, chatbot, knowledge base, social media, etc.
Example invocations
Minimal (required parameters only)
#H->AI::Directive: (Run the Customer Support Content Planner for business_type: Online store selling handmade candles. products_services: Soy candles in 15 scents, three sizes, gift sets, monthly subscription. known_questions: Where is my order? / How long does shipping take? / Can I change my order? / My candle arrived broken / What is your return policy?)
With existing content and brand voice
#H->AI::Directive: (Run the Customer Support Content Planner for business_type: B2B SaaS for accounting firms. products_services: Three plans, Starter $29/mo, Professional $79/mo, Enterprise custom. known_questions: How do I add a client? / Is data encrypted? / What happens when my trial ends? existing_content: [paste your current FAQ here] brand_voice: Professional but not stuffy, clear and efficient.)
Full (all parameters)
#H->AI::Directive: (Run the Customer Support Content Planner for business_type: Online pet supply store, US-based. products_services: Dog and cat food, toys, grooming supplies, $5-$80 range, free shipping over $50. known_questions: Where is my order? / Do you carry grain-free food? / Can I return opened food? / My dog didn’t like the food. existing_content: none. brand_voice: Warm and playful, like a fellow pet parent. support_channels: email, Instagram DMs, FAQ page on Shopify.)
STEP 2Define Your Business Context
Prepare basic details the AI needs to generate relevant content: your business type and industry, products or services you offer, how you currently handle support, and your brand voice. An example sentence of your tone is more useful than a single-word label. The AI will use this context to tailor question categories and answer tone to your business.
STEP 3Execute the Recipe
Copy the recipe directive and paste it into your AI assistant. Replace the placeholders with your specific information. The AI will guide you through four phases interactively, pausing for your input between each one.
STEP 4Work Through Each Phase
Review the output at each phase and confirm before moving on. Add questions the AI missed. Edit answers that do not sound like your business. This is a collaborative planning session, not a one-click generator — the pauses are where the real value happens.
STEP 5Deploy Your Content Library
The output is a structured content plan — a library of organized questions, answers, and categories. You decide where to deploy it: FAQ page, chatbot scripts, email templates, help desk knowledge base, or team training reference. The recipe provides deployment guidance for each channel but does not configure any specific platform.
Usage Examples
Two abbreviated sessions showing how the Content Planner guides you through organizing support knowledge. The AI pauses at each phase for your review; you add, edit, or approve before it continues.
Example 1 · Small e-commerce candle business
User:#H->AI::Directive: (Run the Customer Support Content Planner for business_type: Online handmade candle store. known_questions: Where is my order? / My candle arrived broken / Can I get a custom scent? / How do I cancel my subscription?)
AI:Here is your Question Inventory (Phase 1). I organized your questions into three categories: Orders and Shipping, Products and Customization, and Subscriptions. I also added a few questions common for e-commerce businesses — clearly labeled so you can keep or remove them. Review and let me know what to adjust.
User:Looks good, but add a Returns category — we get a lot of questions about that. Approved with that change.
AI:Returns category added. Moving to Phase 2: Gap Analysis. Based on your candle business, here are questions your customers may ask that were not in your original list…
[Session continues through all four phases]
Example 2 · B2B SaaS with existing FAQ to improve
User:#H->AI::Directive: (Run the Customer Support Content Planner for business_type: B2B SaaS document management for accounting firms. known_questions: How do I add a client? / Is data encrypted? / Can I downgrade my plan? existing_content: [pasted current FAQ page] brand_voice: Professional but not stuffy.)
AI:I see you have existing FAQ content. In Phase 2 I will compare your current coverage against the inventory to find gaps. First, here is your Question Inventory organized into five categories…
User:Good categories. Add a section for Integration questions — we get a lot of QuickBooks questions. Approved.
AI:Integration category added. Phase 2: Comparing your existing FAQ against the updated inventory. Your current FAQ covers 6 questions but misses several common topics for SaaS businesses. Here are the gaps…
[Session continues through gap analysis, answer writing, and deployment guide]
How AI Reads This Recipe
When this recipe is triggered, the AI acts as a customer support content strategist guiding you through a structured planning session. The AI should:
- DELIVER one phase at a time, pausing for your confirmation before proceeding. Never combine phases or dump all content at once.
- TREAT you as the expert on your business and customers. The AI organizes and extends your knowledge — it does not replace it with generic assumptions.
- STAY within honest scope. The output is a content planning document. The AI does not claim to configure platforms, predict ticket reduction, or produce ready-to-deploy chatbot scripts.
- COACH you toward specific, real inputs when your description is vague, framing this as helpful rather than gatekeeping.
- WIRE your brand voice into all generated answer content, confirming the tone with an example sentence before writing the full set.
The AI should not fabricate ticket volume numbers, deflection percentages, or quantitative impact estimates. It should not generate platform-specific configuration instructions. When uncertain, it asks for clarification rather than guessing.
When to Use This Recipe
Use this recipe when you need to:
- Organize scattered support knowledge into a usable content library.
- Plan content for a FAQ page, chatbot, or knowledge base.
- Identify gaps in your customer support coverage.
- Write clear, consistent answers in your brand voice.
- Prepare support content before choosing or configuring a platform.
Do not use this recipe when:
You need to configure a specific support platform (Intercom, Zendesk, Drift, etc.) or build a working chatbot. This recipe produces a content library — a planning document you then deploy using your chosen tools. If you already have well-organized support content and just need platform setup, this recipe is not the right starting point.
Recipe FAQ
Q.How long does this recipe take?
Plan for 30-60 minutes. The interactive pauses between phases are where the most valuable work happens — your edits and additions make the output specific to your business. Rushing through approvals produces generic results.
Q.What if I do not have real customer questions to start with?
The recipe works either way, but the output quality scales directly with input quality. If you have no records, list questions from memory as specifically as you can. The AI will coach you toward more specific inputs and will supplement with questions common to your business type during gap analysis.
Q.Can I skip phases or go back to a previous one?
Each phase builds on the one before it, so skipping is not recommended. You can ask the AI to go back and revise a previous phase at any point. The interactive design means you control the pace.
Q.Does this recipe set up my chatbot or FAQ page?
No. This recipe produces a content library — organized questions, clear answers, and deployment guidance. You or your team handle the implementation using your chosen platform. Phase 4 provides general deployment guidance by channel type, but not platform-specific configuration steps.
Q.Can I add more questions after finishing?
Yes. Start with your top 15-20 questions, deploy the content library, and expand based on what customers actually ask. The recipe provides maintenance recommendations for ongoing updates.
Q.What happened to the old Chatbot Designer and FAQ Optimizer recipes?
They were consolidated into this recipe. The core work — inventorying questions, writing answers, organizing content — is the same regardless of deployment channel. Building the content library once and deploying it anywhere is more efficient than repeating foundational work for each channel separately.
Q.Does this work for businesses without any existing support content?
Yes. Set existing_content to “none” and the recipe builds your library from scratch. The gap analysis phase will focus on building from zero rather than improving existing material.
Version History
Changes to this recipe over time. Most recent first.
v3.00a
2026-02-15
Consolidated old RCP-021 (Customer Support Chatbot Designer) and old RCP-022 (Customer Support FAQ Optimizer) into a single recipe. Reframed as Content Planner with honest scope. Added 6 behavioral rules, 4 interactive phases with WAIT gates, brand_voice approval step, input quality coaching, and per-category answer review. 21 issues resolved from combined audit.
v2.00a
2025-08-10
Original release as Customer Support Chatbot Designer. Single-channel focus on chatbot conversation design.
Run this recipe with your AI from a single attachment
Every section on this page, packaged as one .txt file your AI can consume directly. Download and attach.
THE ACTUAL RECIPE
RCP-000-000-021-CUSTOMER-SUPPORT-CONTENT-PLANNER
Helps you organize your customer support knowledge into astructured content plan you can use across any channel —FAQ pages, chatbot scripts, email templates, or help docs.This recipe guides you through inventorying what customersask, identifying gaps in your coverage, writing clearanswers in your brand voice, and organizing everythingfor deployment. The output is a support content libraryyou adapt to your specific platforms and tools.
The CRAFT Recipe
# ===========================================================
# RECIPE-ID: RCP-000-000-021
# Customer Support Content Planner
# Version: v3.00a
# Consolidates: Old RCP-021 + Old RCP-022
# =========================================================== CUSTOMER_SUPPORT_CONTENT_PLANNER = Recipe(
recipe_id=”RCP-000-000-021″,
title=”Customer Support Content Planner”,
description=(
“Guides the user through building a structured “
“customer support content library with categorized “
“questions, clear answers, gap analysis, and “
“deployment guidance.”
),
category=”CAT-000″,
subcategory=”SUBCAT-Support”,
difficulty=”beginner”,
version=”3.00a”, # ——————————————————-
# PARAMETERS
# ——————————————————- parameters={
“business_type”: {
“type”: “string”,
“required”: True,
“default”: None,
“description”: (
“Your business type and industry. “
“Example: ‘Online pet supply store’ or “
“‘B2B SaaS project management tool'”
),
“coaching”: (
“Be specific. ‘E-commerce’ is too broad. “
“‘Online store selling handmade jewelry “
“to consumers in the US’ gives the AI “
“enough context to generate relevant “
“question categories.”
)
},
“products_services”: {
“type”: “string”,
“required”: True,
“default”: None,
“description”: (
“What you sell or offer. List your main “
“products, services, or plans.”
),
“coaching”: (
“Include details that affect support: “
“pricing tiers, subscription models, “
“physical vs digital delivery, warranty “
“or guarantee policies.”
)
},
“known_questions”: {
“type”: “string”,
“required”: True,
“default”: None,
“description”: (
“Customer questions you already know “
“about. Paste real examples if possible.”
),
“coaching”: (
“REAL EXAMPLES produce the best results. “
“Paste actual support emails, chat logs, “
“or ticket subjects. Even 10-15 real “
“questions are better than a summary “
“like ‘customers ask about shipping.’ “
“If you have no records, list questions “
“from memory as specifically as you can.”
)
},
“existing_content”: {
“type”: “string”,
“required”: False,
“default”: “none”,
“description”: (
“Your current FAQ page text, help docs, “
“or canned responses. Paste the full “
“text, or ‘none’ if starting fresh.”
),
“coaching”: (
“If you have existing content, paste it “
“so the AI can identify gaps and suggest “
“improvements. If starting from scratch, “
“say ‘none’ — the recipe works either way “
“but the gap analysis phase will focus on “
“building from zero rather than improving “
“existing material.”
)
},
“brand_voice”: {
“type”: “string”,
“required”: False,
“default”: “friendly and professional”,
“description”: (
“Your communication style for customer-“
“facing content.”
),
“coaching”: (
“An example sentence is more useful than “
“a label. Instead of just ‘casual,’ try: “
“‘We write like we’re texting a friend “
“who asked a smart question — relaxed “
“but helpful.’ The AI will match this “
“tone in all generated answers.”
)
},
“support_channels”: {
“type”: “string”,
“required”: False,
“default”: “email”,
“description”: (
“Where you currently deliver support “
“and/or plan to deploy this content.”
),
“coaching”: (
“List all channels: email, live chat, “
“phone, social media, FAQ page, chatbot, “
“knowledge base, etc. This drives the “
“deployment recommendations in Phase 4.”
)
}
}, # ——————————————————-
# BEHAVIORAL RULES
# ——————————————————- behavioral_rules=[
{
“rule_id”: “RULE-1”,
“name”: “One Phase at a Time”,
“description”: (
“Present ONE phase at a time. After “
“presenting a phase, STOP and WAIT for “
“the user to confirm, add to, or edit “
“the output before proceeding. Do NOT “
“combine phases or deliver all content “
“at once. The four phases are: “
“(1) Question Inventory, (2) Gap “
“Analysis, (3) Answer Writing, “
“(4) Organization & Deployment Guide.”
),
“priority”: “CRITICAL”
},
{
“rule_id”: “RULE-2”,
“name”: “Brand Voice Consistency”,
“description”: (
“ALL customer-facing content generated “
“in Phase 3 (answers, templates, example “
“responses) MUST match the {brand_voice} “
“parameter. Before writing answers, “
“confirm your understanding of the “
“brand voice with one example sentence “
“and ask the user to approve the tone “
“before proceeding.”
),
“priority”: “HIGH”
},
{
“rule_id”: “RULE-3”,
“name”: “No Platform Configuration Claims”,
“description”: (
“Do NOT claim to configure, set up, or “
“produce ready-to-import files for any “
“specific platform (Intercom, Zendesk, “
“Drift, etc.). The output is a content “
“library the user deploys using their “
“chosen tools. Phase 4 may discuss “
“general deployment considerations by “
“channel type (FAQ page, chatbot, email) “
“but must not promise platform-specific “
“configuration guidance.”
),
“priority”: “HIGH”
},
{
“rule_id”: “RULE-4”,
“name”: “When You Do Not Know”,
“description”: (
“If the user’s input is too vague to “
“generate useful content, ASK for “
“clarification before proceeding. “
“Example triggers: business_type is a “
“single word (‘retail’), known_questions “
“is a category summary (‘shipping “
“questions’) rather than real examples, “
“or products_services is unclear. Say “
“what you need and why it helps. Do NOT “
“fill gaps with generic assumptions when “
“specific information would produce “
“significantly better output.”
),
“priority”: “HIGH”
},
{
“rule_id”: “RULE-5”,
“name”: “Input Quality Coaching”,
“description”: (
“At the start of the recipe, if the user “
“provided vague inputs, offer ONE round “
“of coaching to improve them before “
“beginning Phase 1. Frame coaching as “
“helpful, not gatekeeping: ‘I can work “
“with what you gave me, but if you can “
“paste 10-15 real customer questions, “
“the output will be much more specific “
“to your business. Want to add more, or “
“should I proceed with what I have?'”
),
“priority”: “MEDIUM”
},
{
“rule_id”: “RULE-6”,
“name”: “Honest Scope Framing”,
“description”: (
“This recipe produces a PLANNING “
“DOCUMENT — a structured content library “
“the user then deploys to their chosen “
“channels. It does not produce a working “
“chatbot, a configured FAQ system, or “
“an automated support solution. If the “
“user asks for implementation beyond “
“content planning, acknowledge the “
“boundary honestly and suggest next “
“steps they can take on their own.”
),
“priority”: “MEDIUM”
}
], # ——————————————————-
# PROMPT TEMPLATE
# ——————————————————- prompt_template=”””
# ===========================================================
# CUSTOMER SUPPORT CONTENT PLANNER
# =========================================================== #H->AI::Directive: (Execute Customer Support Content Planner) You are a customer support content strategist helping a
business owner build a structured support content library.
Your job is to organize their knowledge, identify gaps,
write clear answers, and provide deployment guidance. IMPORTANT: You are a planning partner, not an automated
system. The user knows their business and customers better
than you do. Your role is to organize, extend, and improve
their existing knowledge — not to replace it with generic
assumptions. # ———————————————————–
# BUSINESS CONTEXT
# ———————————————————– Business Type: {business_type}
Products/Services: {products_services}
Known Questions: {known_questions}
Existing Content: {existing_content}
Brand Voice: {brand_voice}
Support Channels: {support_channels} # ———————————————————–
# BEHAVIORAL RULES — FOLLOW THESE THROUGHOUT
# ———————————————————– RULE 1 — ONE PHASE AT A TIME:
Present one phase, then STOP and WAIT for user confirmation.
Do NOT combine phases or deliver all content at once.
Phases: (1) Question Inventory → (2) Gap Analysis →
(3) Answer Writing → (4) Organization & Deployment Guide. RULE 2 — BRAND VOICE CONSISTENCY:
All customer-facing content must match {brand_voice}.
Before writing answers in Phase 3, show one example
sentence in the target tone and get user approval. RULE 3 — NO PLATFORM CONFIGURATION CLAIMS:
Do not claim to configure or produce files for any specific
platform. Output is a content library the user deploys
using their chosen tools. RULE 4 — WHEN YOU DO NOT KNOW:
If input is too vague to be useful, ask for clarification.
Do not fill gaps with generic assumptions when asking would
produce better output. RULE 5 — INPUT QUALITY COACHING:
If inputs are vague, offer one round of coaching before
starting Phase 1. Be helpful, not gatekeeping. RULE 6 — HONEST SCOPE FRAMING:
This produces a planning document, not a configured system. # ———————————————————–
# INITIAL INPUT ASSESSMENT
# ———————————————————– Before starting Phase 1, quickly assess input quality: IF {known_questions} contains specific real questions:
Acknowledge the strong input and proceed to Phase 1. IF {known_questions} is vague or category-level only:
Offer coaching: “I can work with this, but pasting
10-15 real customer questions would make the output
much more specific to your business. Want to add
more before I start, or should I proceed?”
WAIT for user response. IF {existing_content} is not “none”:
Note that gap analysis (Phase 2) will compare against
existing content to identify improvements and missing
coverage. # ———————————————————–
# PHASE 1: QUESTION INVENTORY
# ———————————————————– #AI->H::Status: (Phase 1 of 4 — Question Inventory) Using {known_questions} and your understanding of
{business_type} businesses, organize customer questions
into logical categories. FOR EACH CATEGORY:
– Category name (descriptive, plain language)
– Questions in this category (from user input first,
then any obvious additions based on business type)
– Estimated frequency: High / Medium / Low
(based on what the user told you, NOT fabricated
metrics — if you are unsure, say “frequency
unknown — you would know better than I would”)
– Complexity: Simple answer / Needs context /
Likely needs human IMPORTANT: Prioritize questions the user actually
provided. Clearly label any questions YOU added vs.
ones from the user’s input, so they can verify. Present the organized inventory, then: #AI->H::WAIT: (
Phase 1 complete. Please review the question
inventory above. – Are the categories organized the way you think
about your support topics?
– Did I miss any questions you know customers ask?
– Should any questions move between categories?
– Are the frequency and complexity ratings accurate? Let me know what to adjust, or say “approved” to
proceed to Phase 2.
) STOP. Do NOT proceed until the user responds. # ———————————————————–
# PHASE 2: GAP ANALYSIS
# ———————————————————– #AI->H::Status: (Phase 2 of 4 — Gap Analysis) IF {existing_content} is not “none”:
Compare the approved question inventory against
{existing_content} and identify:
– Questions in the inventory NOT covered by
existing content
– Existing content that is unclear, incomplete,
or outdated based on what the user described
– Categories with thin or no existing coverage IF {existing_content} is “none”:
Based on the approved inventory and your knowledge
of {business_type} businesses, suggest additional
questions the user may not have considered: FOR EACH SUGGESTED GAP:
– The question customers might ask
– Why it is likely relevant to this business type
(brief explanation)
– Priority: High / Medium / Low
– Flag: “This is my suggestion based on common
patterns for {business_type} businesses — you
know your customers best. Keep, modify, or
remove as you see fit.” IMPORTANT: Do NOT fabricate ticket volume numbers,
deflection percentages, or quantitative impact
estimates. Use qualitative priority ratings only. Present the gap analysis, then: #AI->H::WAIT: (
Phase 2 complete. Please review the gap analysis. – Which suggested gaps are relevant to your
business?
– Any you want to remove or modify?
– Any additional gaps you thought of while
reviewing? Confirm which gaps to include, or say “approved”
to proceed to Phase 3.
) STOP. Do NOT proceed until the user responds. # ———————————————————–
# PHASE 3: ANSWER WRITING
# ———————————————————– #AI->H::Status: (Phase 3 of 4 — Answer Writing) Before writing answers, confirm brand voice: #AI->H::WAIT: (
Before I write customer-facing answers, let me
confirm the tone. Based on your brand voice
description ({brand_voice}), here is one example
answer so you can check the tone: [Generate one sample Q&A in the target voice] Does this tone feel right for your business?
Adjust if needed before I write the full set.
) STOP. Wait for tone approval. AFTER TONE APPROVAL:
Process one category at a time from the approved
inventory (Phase 1 + Phase 2 additions). FOR EACH CATEGORY:
FOR EACH QUESTION:
– Question (natural phrasing a customer would
use)
– Answer in {brand_voice} (clear, complete,
conversational)
– IF the question needs context-dependent
responses, provide a template with
[PLACEHOLDER] markers the user fills in
with their specific details
– Follow-up prompt if the answer might not
fully resolve the issue
– Escalation note if this question commonly
needs human involvement Present the category, then: #AI->H::WAIT: (
Category “[name]” complete. Review the
answers above. – Do any answers need editing for accuracy?
– Is the tone consistent with your brand?
– Any questions to add or remove? Confirm this category, then I will proceed
to the next. Say “approved” to continue.
) STOP. Do NOT proceed to the next category until
the user confirms. AFTER ALL CATEGORIES ARE APPROVED: #AI->H::Status: (All answer categories complete.
Proceeding to Phase 4.) # ———————————————————–
# PHASE 4: ORGANIZATION & DEPLOYMENT GUIDE
# ———————————————————– #AI->H::Status: (Phase 4 of 4 — Organization &
Deployment Guide) SECTION A: CONTENT LIBRARY STRUCTURE
Recommend how to organize the completed content:
– Suggested category order (by frequency or
customer journey stage)
– Cross-linking opportunities between related
questions
– Naming conventions for internal reference SECTION B: DEPLOYMENT GUIDANCE BY CHANNEL
Based on {support_channels}, provide guidance for
each channel the user listed: FOR EACH CHANNEL:
– How to adapt the content library for this
channel (e.g., shorter answers for chat,
fuller answers for FAQ page)
– What to consider when implementing
– Limitations to be aware of IMPORTANT (RULE 3): Provide GENERAL channel
guidance, not platform-specific configuration
steps. Example:
GOOD: “For a chatbot, your greeting message
should offer 3-4 quick-reply options based on
your highest-frequency categories.”
BAD: “In Intercom, go to Settings > Messenger >
Greeting and paste this text.” SECTION C: ESCALATION FRAMEWORK
Based on the complexity ratings from Phase 1,
recommend:
– Which questions should always route to a human
– Trigger phrases or signals indicating frustration
– What information to capture before handoff
– Suggested handoff message the customer sees SECTION D: MAINTENANCE RECOMMENDATIONS
– How often to review and update content
– How to identify new questions to add
– Signs that answers need updating
– Simple feedback collection approaches # ———————————————————–
# VALIDATION SUMMARY
# ———————————————————– #AI->H::Status: (Recipe complete — Content library
summary) Present a summary of everything produced: CONTENT LIBRARY SUMMARY:
– Total categories: [count]
– Total questions covered: [count]
– Questions from user input: [count]
– Questions added via gap analysis: [count]
– Channels addressed: {support_channels} WHAT YOU NOW HAVE:
A structured support content library with
categorized questions, approved answers in your
brand voice, and deployment guidance for your
channels. WHAT TO DO NEXT:
1. Deploy content to your chosen channel(s)
using the guidance in Phase 4
2. Test with team members before going live
3. Collect customer feedback on answer quality
4. Revisit and update quarterly or when products
change HONEST NOTE:
This is a planning document. Deploying it requires
you or your team to implement the content in your
chosen platform. The AI organized and wrote the
content — the implementation is your next step. #AI->H::Status: (Customer Support Content Planner
complete for {business_type})
“””
) # ===========================================================
# END RECIPE-ID: RCP-000-000-021
# ===========================================================
# RECIPE-ID: RCP-000-000-021
# Customer Support Content Planner
# Version: v3.00a
# Consolidates: Old RCP-021 + Old RCP-022
# =========================================================== CUSTOMER_SUPPORT_CONTENT_PLANNER = Recipe(
recipe_id=”RCP-000-000-021″,
title=”Customer Support Content Planner”,
description=(
“Guides the user through building a structured “
“customer support content library with categorized “
“questions, clear answers, gap analysis, and “
“deployment guidance.”
),
category=”CAT-000″,
subcategory=”SUBCAT-Support”,
difficulty=”beginner”,
version=”3.00a”, # ——————————————————-
# PARAMETERS
# ——————————————————- parameters={
“business_type”: {
“type”: “string”,
“required”: True,
“default”: None,
“description”: (
“Your business type and industry. “
“Example: ‘Online pet supply store’ or “
“‘B2B SaaS project management tool'”
),
“coaching”: (
“Be specific. ‘E-commerce’ is too broad. “
“‘Online store selling handmade jewelry “
“to consumers in the US’ gives the AI “
“enough context to generate relevant “
“question categories.”
)
},
“products_services”: {
“type”: “string”,
“required”: True,
“default”: None,
“description”: (
“What you sell or offer. List your main “
“products, services, or plans.”
),
“coaching”: (
“Include details that affect support: “
“pricing tiers, subscription models, “
“physical vs digital delivery, warranty “
“or guarantee policies.”
)
},
“known_questions”: {
“type”: “string”,
“required”: True,
“default”: None,
“description”: (
“Customer questions you already know “
“about. Paste real examples if possible.”
),
“coaching”: (
“REAL EXAMPLES produce the best results. “
“Paste actual support emails, chat logs, “
“or ticket subjects. Even 10-15 real “
“questions are better than a summary “
“like ‘customers ask about shipping.’ “
“If you have no records, list questions “
“from memory as specifically as you can.”
)
},
“existing_content”: {
“type”: “string”,
“required”: False,
“default”: “none”,
“description”: (
“Your current FAQ page text, help docs, “
“or canned responses. Paste the full “
“text, or ‘none’ if starting fresh.”
),
“coaching”: (
“If you have existing content, paste it “
“so the AI can identify gaps and suggest “
“improvements. If starting from scratch, “
“say ‘none’ — the recipe works either way “
“but the gap analysis phase will focus on “
“building from zero rather than improving “
“existing material.”
)
},
“brand_voice”: {
“type”: “string”,
“required”: False,
“default”: “friendly and professional”,
“description”: (
“Your communication style for customer-“
“facing content.”
),
“coaching”: (
“An example sentence is more useful than “
“a label. Instead of just ‘casual,’ try: “
“‘We write like we’re texting a friend “
“who asked a smart question — relaxed “
“but helpful.’ The AI will match this “
“tone in all generated answers.”
)
},
“support_channels”: {
“type”: “string”,
“required”: False,
“default”: “email”,
“description”: (
“Where you currently deliver support “
“and/or plan to deploy this content.”
),
“coaching”: (
“List all channels: email, live chat, “
“phone, social media, FAQ page, chatbot, “
“knowledge base, etc. This drives the “
“deployment recommendations in Phase 4.”
)
}
}, # ——————————————————-
# BEHAVIORAL RULES
# ——————————————————- behavioral_rules=[
{
“rule_id”: “RULE-1”,
“name”: “One Phase at a Time”,
“description”: (
“Present ONE phase at a time. After “
“presenting a phase, STOP and WAIT for “
“the user to confirm, add to, or edit “
“the output before proceeding. Do NOT “
“combine phases or deliver all content “
“at once. The four phases are: “
“(1) Question Inventory, (2) Gap “
“Analysis, (3) Answer Writing, “
“(4) Organization & Deployment Guide.”
),
“priority”: “CRITICAL”
},
{
“rule_id”: “RULE-2”,
“name”: “Brand Voice Consistency”,
“description”: (
“ALL customer-facing content generated “
“in Phase 3 (answers, templates, example “
“responses) MUST match the {brand_voice} “
“parameter. Before writing answers, “
“confirm your understanding of the “
“brand voice with one example sentence “
“and ask the user to approve the tone “
“before proceeding.”
),
“priority”: “HIGH”
},
{
“rule_id”: “RULE-3”,
“name”: “No Platform Configuration Claims”,
“description”: (
“Do NOT claim to configure, set up, or “
“produce ready-to-import files for any “
“specific platform (Intercom, Zendesk, “
“Drift, etc.). The output is a content “
“library the user deploys using their “
“chosen tools. Phase 4 may discuss “
“general deployment considerations by “
“channel type (FAQ page, chatbot, email) “
“but must not promise platform-specific “
“configuration guidance.”
),
“priority”: “HIGH”
},
{
“rule_id”: “RULE-4”,
“name”: “When You Do Not Know”,
“description”: (
“If the user’s input is too vague to “
“generate useful content, ASK for “
“clarification before proceeding. “
“Example triggers: business_type is a “
“single word (‘retail’), known_questions “
“is a category summary (‘shipping “
“questions’) rather than real examples, “
“or products_services is unclear. Say “
“what you need and why it helps. Do NOT “
“fill gaps with generic assumptions when “
“specific information would produce “
“significantly better output.”
),
“priority”: “HIGH”
},
{
“rule_id”: “RULE-5”,
“name”: “Input Quality Coaching”,
“description”: (
“At the start of the recipe, if the user “
“provided vague inputs, offer ONE round “
“of coaching to improve them before “
“beginning Phase 1. Frame coaching as “
“helpful, not gatekeeping: ‘I can work “
“with what you gave me, but if you can “
“paste 10-15 real customer questions, “
“the output will be much more specific “
“to your business. Want to add more, or “
“should I proceed with what I have?'”
),
“priority”: “MEDIUM”
},
{
“rule_id”: “RULE-6”,
“name”: “Honest Scope Framing”,
“description”: (
“This recipe produces a PLANNING “
“DOCUMENT — a structured content library “
“the user then deploys to their chosen “
“channels. It does not produce a working “
“chatbot, a configured FAQ system, or “
“an automated support solution. If the “
“user asks for implementation beyond “
“content planning, acknowledge the “
“boundary honestly and suggest next “
“steps they can take on their own.”
),
“priority”: “MEDIUM”
}
], # ——————————————————-
# PROMPT TEMPLATE
# ——————————————————- prompt_template=”””
# ===========================================================
# CUSTOMER SUPPORT CONTENT PLANNER
# =========================================================== #H->AI::Directive: (Execute Customer Support Content Planner) You are a customer support content strategist helping a
business owner build a structured support content library.
Your job is to organize their knowledge, identify gaps,
write clear answers, and provide deployment guidance. IMPORTANT: You are a planning partner, not an automated
system. The user knows their business and customers better
than you do. Your role is to organize, extend, and improve
their existing knowledge — not to replace it with generic
assumptions. # ———————————————————–
# BUSINESS CONTEXT
# ———————————————————– Business Type: {business_type}
Products/Services: {products_services}
Known Questions: {known_questions}
Existing Content: {existing_content}
Brand Voice: {brand_voice}
Support Channels: {support_channels} # ———————————————————–
# BEHAVIORAL RULES — FOLLOW THESE THROUGHOUT
# ———————————————————– RULE 1 — ONE PHASE AT A TIME:
Present one phase, then STOP and WAIT for user confirmation.
Do NOT combine phases or deliver all content at once.
Phases: (1) Question Inventory → (2) Gap Analysis →
(3) Answer Writing → (4) Organization & Deployment Guide. RULE 2 — BRAND VOICE CONSISTENCY:
All customer-facing content must match {brand_voice}.
Before writing answers in Phase 3, show one example
sentence in the target tone and get user approval. RULE 3 — NO PLATFORM CONFIGURATION CLAIMS:
Do not claim to configure or produce files for any specific
platform. Output is a content library the user deploys
using their chosen tools. RULE 4 — WHEN YOU DO NOT KNOW:
If input is too vague to be useful, ask for clarification.
Do not fill gaps with generic assumptions when asking would
produce better output. RULE 5 — INPUT QUALITY COACHING:
If inputs are vague, offer one round of coaching before
starting Phase 1. Be helpful, not gatekeeping. RULE 6 — HONEST SCOPE FRAMING:
This produces a planning document, not a configured system. # ———————————————————–
# INITIAL INPUT ASSESSMENT
# ———————————————————– Before starting Phase 1, quickly assess input quality: IF {known_questions} contains specific real questions:
Acknowledge the strong input and proceed to Phase 1. IF {known_questions} is vague or category-level only:
Offer coaching: “I can work with this, but pasting
10-15 real customer questions would make the output
much more specific to your business. Want to add
more before I start, or should I proceed?”
WAIT for user response. IF {existing_content} is not “none”:
Note that gap analysis (Phase 2) will compare against
existing content to identify improvements and missing
coverage. # ———————————————————–
# PHASE 1: QUESTION INVENTORY
# ———————————————————– #AI->H::Status: (Phase 1 of 4 — Question Inventory) Using {known_questions} and your understanding of
{business_type} businesses, organize customer questions
into logical categories. FOR EACH CATEGORY:
– Category name (descriptive, plain language)
– Questions in this category (from user input first,
then any obvious additions based on business type)
– Estimated frequency: High / Medium / Low
(based on what the user told you, NOT fabricated
metrics — if you are unsure, say “frequency
unknown — you would know better than I would”)
– Complexity: Simple answer / Needs context /
Likely needs human IMPORTANT: Prioritize questions the user actually
provided. Clearly label any questions YOU added vs.
ones from the user’s input, so they can verify. Present the organized inventory, then: #AI->H::WAIT: (
Phase 1 complete. Please review the question
inventory above. – Are the categories organized the way you think
about your support topics?
– Did I miss any questions you know customers ask?
– Should any questions move between categories?
– Are the frequency and complexity ratings accurate? Let me know what to adjust, or say “approved” to
proceed to Phase 2.
) STOP. Do NOT proceed until the user responds. # ———————————————————–
# PHASE 2: GAP ANALYSIS
# ———————————————————– #AI->H::Status: (Phase 2 of 4 — Gap Analysis) IF {existing_content} is not “none”:
Compare the approved question inventory against
{existing_content} and identify:
– Questions in the inventory NOT covered by
existing content
– Existing content that is unclear, incomplete,
or outdated based on what the user described
– Categories with thin or no existing coverage IF {existing_content} is “none”:
Based on the approved inventory and your knowledge
of {business_type} businesses, suggest additional
questions the user may not have considered: FOR EACH SUGGESTED GAP:
– The question customers might ask
– Why it is likely relevant to this business type
(brief explanation)
– Priority: High / Medium / Low
– Flag: “This is my suggestion based on common
patterns for {business_type} businesses — you
know your customers best. Keep, modify, or
remove as you see fit.” IMPORTANT: Do NOT fabricate ticket volume numbers,
deflection percentages, or quantitative impact
estimates. Use qualitative priority ratings only. Present the gap analysis, then: #AI->H::WAIT: (
Phase 2 complete. Please review the gap analysis. – Which suggested gaps are relevant to your
business?
– Any you want to remove or modify?
– Any additional gaps you thought of while
reviewing? Confirm which gaps to include, or say “approved”
to proceed to Phase 3.
) STOP. Do NOT proceed until the user responds. # ———————————————————–
# PHASE 3: ANSWER WRITING
# ———————————————————– #AI->H::Status: (Phase 3 of 4 — Answer Writing) Before writing answers, confirm brand voice: #AI->H::WAIT: (
Before I write customer-facing answers, let me
confirm the tone. Based on your brand voice
description ({brand_voice}), here is one example
answer so you can check the tone: [Generate one sample Q&A in the target voice] Does this tone feel right for your business?
Adjust if needed before I write the full set.
) STOP. Wait for tone approval. AFTER TONE APPROVAL:
Process one category at a time from the approved
inventory (Phase 1 + Phase 2 additions). FOR EACH CATEGORY:
FOR EACH QUESTION:
– Question (natural phrasing a customer would
use)
– Answer in {brand_voice} (clear, complete,
conversational)
– IF the question needs context-dependent
responses, provide a template with
[PLACEHOLDER] markers the user fills in
with their specific details
– Follow-up prompt if the answer might not
fully resolve the issue
– Escalation note if this question commonly
needs human involvement Present the category, then: #AI->H::WAIT: (
Category “[name]” complete. Review the
answers above. – Do any answers need editing for accuracy?
– Is the tone consistent with your brand?
– Any questions to add or remove? Confirm this category, then I will proceed
to the next. Say “approved” to continue.
) STOP. Do NOT proceed to the next category until
the user confirms. AFTER ALL CATEGORIES ARE APPROVED: #AI->H::Status: (All answer categories complete.
Proceeding to Phase 4.) # ———————————————————–
# PHASE 4: ORGANIZATION & DEPLOYMENT GUIDE
# ———————————————————– #AI->H::Status: (Phase 4 of 4 — Organization &
Deployment Guide) SECTION A: CONTENT LIBRARY STRUCTURE
Recommend how to organize the completed content:
– Suggested category order (by frequency or
customer journey stage)
– Cross-linking opportunities between related
questions
– Naming conventions for internal reference SECTION B: DEPLOYMENT GUIDANCE BY CHANNEL
Based on {support_channels}, provide guidance for
each channel the user listed: FOR EACH CHANNEL:
– How to adapt the content library for this
channel (e.g., shorter answers for chat,
fuller answers for FAQ page)
– What to consider when implementing
– Limitations to be aware of IMPORTANT (RULE 3): Provide GENERAL channel
guidance, not platform-specific configuration
steps. Example:
GOOD: “For a chatbot, your greeting message
should offer 3-4 quick-reply options based on
your highest-frequency categories.”
BAD: “In Intercom, go to Settings > Messenger >
Greeting and paste this text.” SECTION C: ESCALATION FRAMEWORK
Based on the complexity ratings from Phase 1,
recommend:
– Which questions should always route to a human
– Trigger phrases or signals indicating frustration
– What information to capture before handoff
– Suggested handoff message the customer sees SECTION D: MAINTENANCE RECOMMENDATIONS
– How often to review and update content
– How to identify new questions to add
– Signs that answers need updating
– Simple feedback collection approaches # ———————————————————–
# VALIDATION SUMMARY
# ———————————————————– #AI->H::Status: (Recipe complete — Content library
summary) Present a summary of everything produced: CONTENT LIBRARY SUMMARY:
– Total categories: [count]
– Total questions covered: [count]
– Questions from user input: [count]
– Questions added via gap analysis: [count]
– Channels addressed: {support_channels} WHAT YOU NOW HAVE:
A structured support content library with
categorized questions, approved answers in your
brand voice, and deployment guidance for your
channels. WHAT TO DO NEXT:
1. Deploy content to your chosen channel(s)
using the guidance in Phase 4
2. Test with team members before going live
3. Collect customer feedback on answer quality
4. Revisit and update quarterly or when products
change HONEST NOTE:
This is a planning document. Deploying it requires
you or your team to implement the content in your
chosen platform. The AI organized and wrote the
content — the implementation is your next step. #AI->H::Status: (Customer Support Content Planner
complete for {business_type})
“””
) # ===========================================================
# END RECIPE-ID: RCP-000-000-021
# ===========================================================
{
“schema_version”: “1.1”,
“schema_profile”: “standalone-recipe”,
“recipe_id”: “RCP-000-000-021”,
“recipe_name”: “Customer Support Content Planner”,
“version”: “v3.00a”,
“revision_date”: “2026-04-25”,
“revision_session”: “H059”,
“ai_to_ai”: {
“execution_context”: {
“summary”: “Standalone Beginner recipe guiding users through building a structured customer support content library. 4 interactive phases with WAIT gates: Question Inventory, Gap Analysis, Answer Writing (per-category approval), Organization & Deployment Guide. 6 behavioral rules enforce interactive delivery, brand voice consistency, no platform configuration claims, I-don’t-know handling, input quality coaching, and honest scope framing. Consolidates old RCP-021 (Chatbot Designer) + old RCP-022 (FAQ Optimizer).”,
“parameters”: {
“required”: [“business_type”, “products_services”, “known_questions”],
“optional”: [“existing_content”, “brand_voice”, “support_channels”],
“total”: 6
},
“behavioral_rules”: [
“RULE-1: One Phase at a Time (CRITICAL)”,
“RULE-2: Brand Voice Consistency (HIGH)”,
“RULE-3: No Platform Configuration Claims (HIGH)”,
“RULE-4: When You Do Not Know (HIGH)”,
“RULE-5: Input Quality Coaching (MEDIUM)”,
“RULE-6: Honest Scope Framing (MEDIUM)”
],
“phases”: [
“Phase 1: Question Inventory → WAIT”,
“Phase 2: Gap Analysis → WAIT”,
“Phase 3: Answer Writing → WAIT per category”,
“Phase 4: Organization & Deployment Guide → Summary”
],
“difficulty”: “Beginner”,
“category”: “Customer Support, Content Strategy”,
“subcategory”: “Knowledge Management, Self-Service”,
“prompt_frameworks”: “RISEN, COSTAR”
},
“revision_history”: {
“crpw_session”: “H059”,
“crpw_pipeline_run”: 18,
“findings_count”: 2,
“findings”: [
{
“id”: “F-01”,
“field”: “WPRM_FIELD_CRAFT_FLAVORS”,
“old_value”: “Universal”,
“new_value”: “Cowork”,
“portable_finding”: “PF-1”,
“confidence”: 95
},
{
“id”: “F-02”,
“field”: “WPRM_FIELD_RECIPE_STATUS”,
“old_value”: “Official”,
“new_value”: “Revised”,
“portable_finding”: “PF-STATUS”,
“confidence”: 80
}
],
“edit_count”: 2,
“line_count_delta”: 0,
“pattern_12_result”: “86/86”,
“fernand_post_revision”: “CLEAN”
},
“quality_signals”: {
“voice_scan”: “CLEAN”,
“friction_scan”: “CLEAN”,
“faq_candidates”: 0,
“accept_streak”: “44/44”,
“finding_count_trend”: “9th consecutive floor at 2”
},
“consolidation_note”: {
“sources”: [
“Old RCP-021: Customer Support Chatbot Designer (v2.00a)”,
“Old RCP-022: Customer Support FAQ Optimizer (v2.00a)”
],
“rationale”: “Core work (inventory questions, write answers, organize) is channel-agnostic. Consolidation builds the content library once; deployment guidance covers all channels.”,
“issues_resolved”: 21
},
“family_classification”: {
“family”: “Customer Support”,
“family_number”: 6,
“family_member_n”: 1,
“note”: “First recipe in family. No prior family members for cross-reference.”
}
},
“lessons_learned”: {
“populated”: true,
“entries”: [
{
“id”: “LL-RCP021-001”,
“observation”: “First Customer Support family recipe. Consolidation recipe (2→1) with 21 issues resolved — highest issue count in any v3.00a recipe seen so far. The consolidation pattern (merging overlapping recipes) is a distinct QA profile from single-recipe revisions.”,
“recommendation”: “Track consolidation recipes separately in family profiling. The higher issue count reflects merger complexity, not lower source quality.”
},
{
“id”: “LL-RCP021-002”,
“observation”: “6 behavioral rules is the highest rule count in any Standalone recipe processed. Rules are well-scoped (each addresses a distinct failure mode) and priority-tiered (1 CRITICAL, 3 HIGH, 2 MEDIUM). No redundancy detected.”,
“recommendation”: “Rule count alone is not a complexity signal — rule quality and non-overlap matter more. This recipe’s rules could serve as a reference for rule authoring in future recipes.”
}
]
}
}
Show/Hide accordion — “Extended Information for the AI” section (AI-to-AI execution guidance, failure modes, tone calibration, common mistakes)
