Mine Pharma Complaints: CAPA Alerts with Claude & Slack

Spot unusual complaint trends and trigger CAPA alerts automatically with Mine Pharma + Claude + Slack

Mine Pharma Complaints: CAPA Alerts with Claude & Slack is an n8n automation workflow that monitors newly logged complaint narratives, uses Anthropic Claude to extract structured defect signals, detects statistically unusual spikes, and then escalates via a CAPA ticket and posts a Slack alert when anomalies are found.

What this workflow does

• Polls a complaint-source API on a schedule (or on-demand) to pull recently logged complaints.
• Skips already processed complaints using workflow static data.
• If there are no new complaints, it ends the run without calling Claude.
• Uses Claude to structure complaint narratives, extracting product, batch/lot, defect category, and severity—then applies safe default values if parsing fails.
• Clusters complaints by product and defect category, aggregating signals into cluster volumes.
• Detects anomalies by comparing each cluster’s volume against a rolling historical baseline using a z-score.
• For flagged clusters, it creates a CAPA escalation via an HTTP API and posts to Slack.
• Logs every cluster (escalated and non-escalated) to a trend-store API for ongoing tracking.

Use cases

• Quality teams want rapid CAPA initiation when complaint volumes spike for specific products and defect categories.
• Regulatory/compliance operations need traceable trend logging to support ongoing surveillance.
• Automation engineers can operationalize AI-assisted defect extraction with Claude and route outcomes to Slack for real-time triage.

Technical details

• Integrations (via HTTP Request steps): complaint-source API, CAPA ticketing API, and trend-store API (all using configured HTTP header auth).
• AI: Anthropic Claude to extract product, batch/lot, defect category, and severity.
• n8n nodes/logic: if, set, code, merge, and no op, plus Slack for alerts.