Skip to content

Productivity & Team Ops

Knowledge Base Auto-Updater

Answers found in support threads get drafted as new knowledge base articles.

Saves about ~2 hrs per week

What it does

  • Scans support threads for repeated questions
  • Drafts knowledge base articles from good answers
  • Flags articles that need a human check
  • Keeps your KB current without a dedicated writer

How it works

  1. Support threads scanned
  2. Repeated questions detected
  3. Article drafts generated
  4. Drafts queued for review

What's included

  • Thread scanning
  • Repeat detection
  • Article drafting

Set it up yourself

Follow these 4 steps in order. You will need accounts for: Notion.

Step 1: Connect the support source

  1. Choose the source: your helpdesk API (Gorgias/Zendesk) or the shared support inbox.
  2. Connect it to the workflow with read access.
  3. Pull closed/resolved threads.
  4. Test the pull returns thread text.
  5. Keep the source scoped to resolved threads.

Watch out: Only scan resolved threads — unanswered questions are noise for KB drafting; resolved ones carry the 'working answer'.

Step 2: Set the repeat rule

  1. Define 'repeated': the same question asked 3+ times (by topic or duplicate detection).
  2. Compute a question fingerprint per thread (topic + keywords).
  3. Count occurrences per fingerprint.
  4. When count ≥ 3, mark for drafting.
  5. Test on real thread history.

Watch out: Fingerprint counting is what finds the true repeat questions — asking the AI to 'judge repeat questions' by reading everything is slower and less consistent.

Step 3: Add the draft prompt

  1. Build: 'Draft a knowledge base article from these similar support threads. Structure: title, short answer, step-by-step walkthrough, related links. Use the best answer's wording as the base. Flag any part that needs human confirmation.'
  2. Instruct to stay within the answers' substance.
  3. Paste with API key.
  4. Test on 3 repeated topics.
  5. Check you'd be happy shipping the drafts.

Watch out: Flag-unknowns is critical — a KB article with a confidently-wrong policy answer is worse than no article; the flag forces a human check.

Step 4: Route to review

  1. Create drafts in the KB tool with Status = Review.
  2. Queue the flagged items together.
  3. Review weekly: approve, edit, or reject.
  4. On approval, publish and note the date.
  5. Track KB growth vs ticket reductions.

Watch out: The weekly review cadence is what keeps the KB trustworthy — a queue nobody reviews becomes a pile of unchecked drafts that rot.

Frequently asked questions

What does the Knowledge Base Auto-Updater automation do?

Answers found in support threads get drafted as new knowledge base articles. Scans support threads for repeated questions

How much time does Knowledge Base Auto-Updater save?

It is estimated to save ~2 hrs per week. Actual savings depend on your volume.

Which tools does Knowledge Base Auto-Updater use?

It works with Notion.

What does Knowledge Base Auto-Updater include?

Thread scanning, Repeat detection, Article drafting.

Can I set up Knowledge Base Auto-Updater myself?

Yes. The do-it-yourself guide on this page walks through 4 steps, from getting access to testing the finished automation.