OpenAI Support Bot for n8n: Confidence Slack Handoff

OpenAI Support Bot for n8n: Confidence Slack Handoff

Turn customer questions into accurate answers—only when your n8n workflow can prove it. This OpenAI-powered support bot responds with cited knowledge-base facts, and if confidence is low or citations are missing, it cleanly hands off to your team via Slack instead of guessing.

What this workflow does

• Receives questions by webhook from sources like site chat, a form, an email parser, or a WhatsApp bridge.

• Answers from your knowledge base with citations by requiring the LLM to use numbered FAQ facts and cite them in its response.

• Validates citations and confidence: code checks that citations exist and that the model’s confidence meets your minConfidence threshold (default 0.7).

• Slack handoff when unsure: if the answer is uncited or below threshold, the customer gets a polite handoff message, and your team receives a Slack ticket containing the question, category, and the bot’s draft.

• Logs answered questions with category and cited facts so you can identify gaps and continuously improve the FAQ/KB.

Use cases

• Answering SaaS support questions (e.g., refunds after 10 days) with verified FAQ citations.
• Reducing hallucinations by routing uncertain cases to Slack for human review.
• Building a measurable knowledge-base improvement loop by reviewing which categories were answered and which facts were missing.

Technical details

• n8n nodes/workflow tools: webhook, if, code, http request, respond to webhook, plus sticky note for guidance.

• OpenAI integration: configured via Config + Knowledge Base in the LLM node, using an OpenAI-compatible API (baseUrl, model).

• Slack integration: uses a Slack incoming-webhook in Create Ticket / Alert.

• Setup & testing: send a POST request to the webhook test URL at path support-agent with a JSON body containing question and email.