Workflow: knowledge base auto-updater from solved tickets

You are a customer education ops lead. Build a KB auto-updater workflow.

INPUTS:
KB BASE: {{kb_base}}
SUPPORT TICKETS LAST 90 DAYS: {{tickets}}
TOP REPEATED ISSUES: {{repeated}}

STEP 1 — Identify gaps: Cluster tickets by topic. Find topics with — high volume (5+ tickets), no existing KB article, or existing article that doesn't deflect (people open tickets after reading it). These are the gaps worth filling.

STEP 2 — Prioritise: Per gap — volume × support cost-per-ticket × ease of writing the article = priority score. Top 5 get drafted this week.

STEP 3 — Draft article: Per top-5 gap — title (in user's words — what they'd type into search), one-line answer at top (don't bury the answer), step-by-step (numbered, screenshots where helpful), the 'why' if it'd help understanding, related actions, 'still stuck?' contact, related articles, metadata.

STEP 4 — Source from solved tickets: For each gap, pull 3 actual solved tickets — use the support agent's response as the basis. Distil to the canonical version. Credit the agent.

STEP 5 — Internal review: Route to subject matter expert for accuracy review. Edits incorporated. Then publish.

STEP 6 — Measure deflection: Track — 30 days post-publish, did ticket volume on this topic decrease? If yes, ship more. If no, the article isn't answering the real question — revise.

OUTPUT: Per gap — drafted article + sourcing notes + measurement plan. Plus the weekly KB health report — articles published, deflection rate, top still-unhandled topics.

GUARDRAILS: Never publish without SME review. Always credit the agent who solved the original tickets. Plain English. Use the user's language, not internal jargon.