Workflow Nodes

  1. Start: receives customer message + conversation history
  2. Intent Classifier (LLM): labels as "product_question", "pricing", "demo_request", "complaint", or "other"
  3. Knowledge Retrieval: queries your product/pricing knowledge base (top 3 chunks)
  4. Response Drafter (LLM): generates a helpful, on-brand reply using retrieved context
  5. Conditional Branch: if intent == "demo_request" or confidence < 0.6, route to human handoff
  6. Human Handoff: posts to Slack/CRM and returns a "transferring you to an agent" message
  7. End: returns final response to the chat widget

Setup Steps

  • Import this workflow JSON into Dify
  • Connect your LLM provider (OpenAI / Anthropic / local)
  • Upload product docs, pricing sheets, and FAQs to the knowledge base
  • Configure the human-handoff webhook (Slack / HubSpot / Salesforce)
  • Publish and embed the chat widget on your site

FAQ

What is a Dify sales assistant workflow?

A Dify AI application that handles inbound sales inquiries: classifies intent, retrieves product info from a knowledge base, drafts responses, and can hand off to a human. This template wires an LLM node, knowledge retrieval, and conditional branching.

Do I need coding to deploy this?

No. Dify provides a visual orchestration canvas. You import this workflow JSON, connect your LLM provider key, and upload your product knowledge base. No code required.

All Workflows