The Knowledge Chain

Description

GOAL
Build a workflow that analyzes support tickets weekly, clusters frequent questions, and creates finished FAQ drafts in the review area of the knowledge base, including a notification to the content team.

Adapt the structure to the concepts of your automation tool (for example nodes, steps, zaps, or agent actions). While doing so, preserve the trigger logic, the sequence, and the data flow.

TRIGGER
Type: scheduled, weekly, for example every Friday at 5:00 p.m.
Scope: resolved or closed tickets from the past seven days from the ticketing system (for example Zendesk or Freshdesk).

STEPS (sequential)
1. Retrieve tickets: all resolved or closed tickets from the past seven days, with subject, description, category and resolution.
2. AI step: topic clustering. Use exactly the working prompt "Ticket Analysis and Clustering" (see the Working Prompts section). Output is JSON with the field frequent_topics.
3. AI step: create FAQ drafts. Use exactly the working prompt "FAQ Creation", with the JSON from step 2 as input. Output is JSON with the field faq_entries.
4. Duplicate check: check the drafts against existing entries in the knowledge base. Discard duplicates or flag them as an update.
5. Save: file all remaining drafts in the review area of the knowledge base (Notion, Confluence or GitBook), including tags and priority.
6. Output: notify the content team in Slack or Teams, with a link to the review area and a topic summary including frequencies.

PROMPT SPECIFICATIONS
There are two fixed working prompts for this workflow, see the Working Prompts section. Take both over unchanged and in full into step 2 and step 3.

DATA FLOW
Step 1 to step 2: the retrieved tickets from the past seven days.
Step 2 to step 3: the JSON field frequent_topics, with topic, frequency, relevance and example questions.
Step 3 to steps 4 and 5: the JSON field faq_entries, with title, question, answer, tags and priority.
Step 5 to step 6: the filed draft links and the topic list with frequencies.

ERROR HANDLING
No tickets found: end the run with a notice instead of empty drafts, and check the period, status and category filters.
Topic below the defined minimum frequency: discard it, one-off cases do not belong in the knowledge base.
AI returns no valid JSON: retry the step once, then abort the run and report the error.
Draft is a duplicate of an existing entry: always check every draft against the current knowledge base before saving.

PLACEHOLDER CONVENTION
All placeholders in square brackets in the working prompts are variables. The ticket data and the JSON field frequent_topics are mapped dynamically, as described in the data flow. Minimum frequency, product area, tone of voice, target audience, answer length and documentation URL are not dynamic: set these yourself before activation. List all placeholders that require manual configuration separately at the end.

SELF CHECK BEFORE COMPLETION
Check, before you report the workflow as finished:
1. Are all six steps connected correctly and in the right order?
2. Are both working prompts inserted completely and unchanged, and is the data flow between them mapped correctly?
3. Is the JSON output of both AI steps parsed correctly, and does the duplicate check apply before saving?
4. Are all points listed that require manual configuration: ticketing system access, knowledge base area, team chat channel, LLM access, minimum frequency, product area, tone of voice, target audience, answer length and documentation URL?

Working prompts

You are a support analyst. Analyze the following support tickets from the past week and identify the most frequent questions and problems.

TICKETS:
[TICKET_DATA]

TASK:
1. Group similar questions and problems.
2. Identify the five to ten most frequent topics.
3. Count how often each topic occurs.
4. Rate the relevance for an FAQ on a scale from 1 to 10.

OUTPUT FORMAT (JSON):
{
  "frequent_topics": [
    {
      "topic": "Short description of the topic",
      "frequency": count,
      "relevance": rating from 1 to 10,
      "example_questions": ["Question 1", "Question 2"]
    }
  ]
}

CRITERIA:
At least [MIN_FREQUENCY] occurrences per topic.
Focus on [PRODUCT_AREA], for example billing, features or technical issues.
Individual one-off cases are ignored.

You are a technical writer. Create professional FAQ entries based on the following frequent support questions.

FREQUENT QUESTIONS:
[FREQUENT_TOPICS]

TASK:
Create a complete FAQ entry for each topic with:
1. A concise title, at most 60 characters
2. A clearly phrased question, the way customers would actually ask it
3. An understandable, step-by-step answer
4. Relevant tags for categorization

OUTPUT FORMAT (JSON):
{
  "faq_entries": [
    {
      "title": "...",
      "question": "...",
      "answer": "...",
      "tags": ["tag1", "tag2"],
      "priority": "high/medium/low"
    }
  ]
}

STYLE:
Tone: [BRAND_VOICE]
Target audience: [TARGET_AUDIENCE]
Answer length: [ANSWER_LENGTH]
Avoid jargon and complicated phrasing.
Use clear steps, examples and, where useful, links to [DOCUMENTATION_URL].

JSON export