AI Scam Message Triage in n8n: Claude + Google Sheets Alerts

Stop AI-detected scam messages before they cost you money—automate triage in n8n with Claude + Google Sheets alerts

This n8n workflow receives suspicious calls/messages via an HTTP webhook, analyzes them with Anthropic Claude (plus heuristic keyword scoring), produces calm plain-language safety guidance, logs every check to Google Sheets, and—when risk is high—alerts a trusted family contact via SendGrid.

What this workflow does

• Ingests suspicious content through an n8n Webhook (with sensible defaults for testing when fields are missing).
• Runs heuristic risk checks by scanning message text for common scam patterns such as urgency, payment methods, secrecy, threats, and impersonated authority.
• Assesses likelihood with Anthropic Claude, requesting scam classification, red flags, and an AI risk score/confidence.
• Handles uncertainty safely: if Claude confidence is too low, the workflow stops and asks for more details instead of guessing.
• Calculates a composite 0–100 risk score by blending the heuristic score, Claude’s AI risk score, and the number of AI-identified red flags; assigns a risk tier.
• Generates next steps in simple terms, including calm guidance and, when needed, a family alert message.
• Logs activity in Google Sheets by appending a summary row via the Google Sheets API.
• Escalates when required: sends a SendGrid email to a trusted family contact only for high-risk cases; otherwise returns guidance directly in the webhook response.

Use cases

• Triaging suspicious WhatsApp messages or SMS texts with automated risk scoring and safety instructions.
• Screening potentially fraudulent emails and suspicious calling attempts before replying or paying.
• Building a lightweight fraud “watch” for SaaS operators or families who want documented alerts and response history.

Technical details

• Implemented with core n8n nodes such as Webhook, if, set, code, and wait (plus notes/sticky note documentation).
• Integrates with Anthropic Claude using an Anthropic API credential.
• Writes each triage result to Google Sheets via the Google Sheets API.
• Sends escalation emails using SendGrid.