Stop answering the same questions repeatedly. This workflow identifies resolved Zendesk tickets that could help other customers, uses AI to generate well-structured knowledge base articles, and saves drafts to Notion for your team to review and publish.
Detect resolved ticket in Zendesk
The workflow triggers when a ticket is marked as resolved in Zendesk. It captures the full ticket thread including the original question, troubleshooting steps, and final resolution. Tickets can be filtered by tag or category to focus on documentation-worthy issues.
Evaluate KB article potential with Claude
Using Claude, the workflow evaluates whether the resolved ticket would make a good knowledge base article. It considers factors like question generalizability, solution clarity, and whether similar content already exists. Not every ticket needs an article.
Generate KB article draft with AI
For tickets that pass the evaluation, the AI generates a complete knowledge base article. The draft includes a clear title, problem statement, step-by-step solution, troubleshooting tips, and related topics. The tone is adjusted from conversational support to clear documentation.
Save article draft to Notion
The generated article draft is saved to a designated Notion database with metadata including source ticket, category, and suggested keywords. Your documentation team can review, edit, and publish to your help center from Notion.
Why automate knowledge base creation with AI?
Support teams solve the same problems repeatedly, but rarely have time to document solutions. The knowledge stays locked in ticket histories instead of being available for self-service. AI automation turns resolved tickets into searchable documentation.
Reduce ticket volume with better self-service
Every article you publish deflects future tickets. Customers find answers themselves, and your team spends less time on repetitive questions. Most teams see 15-30% ticket reduction after building out their KB.
Capture knowledge before it's lost
When experienced agents leave, their expertise often goes with them. Automated article generation ensures solutions are documented while the context is fresh, building organizational knowledge over time.
Keep documentation current without dedicated effort
Products change and new issues emerge constantly. By generating articles from recent tickets, your KB stays current with real customer problems rather than becoming outdated.
How to set up automated KB article generation
Setting up this Zendesk to Notion workflow takes about 15 minutes. You'll configure ticket filters, article templates, and the Notion destination.
What you need to get started
- Zendesk account with resolved ticket access
- Claude API key for article generation
- Notion workspace for draft storage
- Existing KB categories or structure defined
Configuring ticket selection criteria
- Filter by ticket tags to focus on common issues
- Exclude tickets that are too account-specific
- Set minimum ticket quality thresholds (clear resolution, customer confirmed)
- Consider filtering by product area or feature
Customizing article format
- Define your KB article structure (intro, steps, troubleshooting)
- Specify tone and style guidelines for your documentation
- Set up category mapping from Zendesk tags to KB sections
- Configure metadata to include (keywords, related articles)
Frequently asked questions about AI article generation
Will the AI articles be ready to publish immediately?
Articles are saved as drafts for human review. While the AI produces high-quality content, your team should verify accuracy and add any missing context before publishing.
How do you prevent duplicate articles from being created?
The AI evaluation step checks for similar existing content and skips tickets that would duplicate existing articles. You can also configure deduplication rules based on titles or categories.
Can I use this with a different KB platform like Intercom or Helpscout?
Yes, you can swap Notion for any documentation platform with API access. The AI generation works the same regardless of where articles are stored.
What if the ticket resolution was incomplete or incorrect?
The AI evaluates ticket quality and skips tickets without clear resolutions. You can also require specific tags (like "documented solution") before triggering article generation.