Step01
Trigger Analysis on Conversation Completion
Trigger when a support ticket is resolved or a chatbot conversation closes. Capture the full transcript, resolution status, product area, source link, and relevant tags.
AdvancedOperations
Build a Self-Improving Customer Feedback Knowledge Base
Keep a support knowledge base current by mining resolved conversations for reusable answers, checking for existing coverage, and publishing only approved, source-linked FAQ updates.
Source episode · 25:05
How I AI: Reid Robinson's Zapier Workflows for CRM Automation, Meeting Prep, and Feedback Loopswith Claire Vo
Reid analyzes each resolved support conversation for its core question and solution, checks the existing knowledge base, routes a proposed FAQ through human review, and adds approved entries to the chatbot source.
Before you start
What you need
- Resolved support tickets or chatbot transcripts
- An existing knowledge base index
- Zapier Tables or another review queue
- An owner for FAQ approval
What you’ll make
An approved FAQ entry linked to its source conversation and available to the internal or customer-facing assistant.
Tools used
- ZapierVisitZapier
Workflow automation platform
Step by step
The workflow
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
5 steps
Step02
Extract Question-and-Answer Pair with AI
Extract the reusable user question and the solution actually confirmed in the conversation. Return unresolved when the transcript lacks a verified answer.
Example prompt
Analyze this resolved support conversation. Return JSON with reusable_question, verified_answer, product_area, prerequisites, source_evidence, resolution_status, and sensitive_details_to_remove. Do not infer a solution that was not confirmed in the transcript.
Step03
Check Knowledge Base and Propose New Entry
Search the current knowledge base semantically and by product area. Update a matching article when the answer adds useful detail; otherwise draft a concise new FAQ with source provenance.
Example prompt
Compare this verified question and answer with the retrieved knowledge entries. Return duplicate, update_existing, or new_entry with the matching article ID and reason. If an update or new entry is warranted, draft concise customer-safe copy without account-specific details.
Step04
Implement a Human Review Step
Send the proposal to a named reviewer with the source conversation, duplicate analysis, and suggested destination. Allow edit, approve, reject, or route back for missing evidence.
Review is for accuracy and reuse, not a ceremonial approval. Give the reviewer the source and the proposed change together.
Step05
Close the Loop by Updating the Knowledge Base
On approval, write the entry to the correct knowledge source, preserve owner and provenance, refresh the chatbot index, and record the published article ID in the review item.
What good looks like
- The extracted answer reflects the actual resolution in the conversation.
- The workflow searches for semantic duplicates before proposing an entry.
- A named reviewer can edit, approve, or reject the draft.
- Approved content reaches the correct knowledge source with provenance.
Build your next product with ChatPRD
Turn an idea into a PRD, user stories, and a plan.
After the steps
Runbook notes
How to recover when the loop fails and where human judgment helps.
Recover
If it goes sideways
- The conversation ended without a verified solution
- Mark it unresolved and route it to support analysis rather than drafting an FAQ.
- A differently worded question creates duplicate content
- Use semantic search plus product area and intent, then update an existing entry when the answer is materially the same.
- The draft exposes customer or account details
- Remove names, identifiers, secrets, and account-specific configuration before the review queue.
- An approved answer later becomes incorrect
- Store owner, product area, source, and review date, then recheck entries when related tickets or docs change.
Keep building
Related workflows
advancedContentChatGPT
Automate UGC Video Editing From Raw Clips with ChatGPT
Use ChatGPT's multi-modal capabilities to automatically edit user-generated content (UGC) style videos. The AI can transcribe clips, select the best takes, assemble a narrative, and redact sensitive on-screen information.
Open
intermediateMarketingChatGPT
Build and Deploy a Live Website with a Single ChatGPT Prompt
use ChatGPT Sites to generate, build, and deploy a fully functional, data-backed website on a `chatgpt.site` domain using a single detailed prompt.
Open
advancedPersonalChatGPT
Use ChatGPT Voice as a Personal Assistant for Logistics and Travel
Use ChatGPT Voice with screen-aware context to delegate and parallel-process complex logistical tasks like booking travel and filing expense reports, all through a single conversation.
Open
Start shipping
better products.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.
Free to start
No credit card
SOC 2 certified
Enterprise ready