Llama 3.2 Chat Support n8n Workflow: Sheets & Slack Escalation

Turn chat into instant answers—and hand off the hard cases

This Llama 3.2 Chat Support n8n Workflow: Sheets & Slack Escalation automatically handles customer chat for an online shoe store. It looks up products and orders in Google Sheets, answers policy questions with Llama 3.2 via Ollama, logs every conversation, and escalates complaints or uncertain cases to Slack for human follow-up.

What this workflow does

• Receives chat messages (or can run on a daily schedule for reporting) and loads store configuration.
• Classifies intent using keyword rules, extracting order numbers and email addresses. If the message is unclear, it uses Llama 3.2 via Ollama to help interpret the request.
• Answers product questions by searching the Products tab in Google Sheets and returning matching items with prices, sizes, and colors.
• Answers order-status questions by looking up the Orders tab in Google Sheets, including flags for delayed or unclear statuses to trigger human follow-up.
• Handles FAQ and policy questions by sending the customer message plus store policy text to Llama 3.2 via Ollama. If the model returns NOT_SURE, the workflow escalates.
• Creates human handoff tickets in the Tickets tab and posts alerts to a Slack channel for complaints, requests, emails, unclear messages, and flagged orders.
• Logs every interaction to the Conversations tab and returns the final reply to the chat. Each morning, it posts a 24-hour activity summary to Slack.

Use cases

• Customers ask for shoe availability by size/color → search Products and reply instantly.
• Customers request order status → look up Orders and flag delays/uncertainty.
• Customers ask about returns/shipping policy → generate answers via Llama 3.2 and escalate when NOT_SURE.
• Complaints, human requests, or unclear messages → ticket + Slack alert.

Technical details

• Google Sheets: Products, Orders, Tickets, Conversations (full logging + handoffs)
• Slack: real-time escalation alerts and daily 24-hour summary
• Ollama + Llama 3.2: intent help when keyword rules are unclear; policy/FAQ answers when needed
• Core n8n building blocks: if, set, code, switch, sticky note