AI-powered WhatsApp customer support for Shopify brands with LLM agents

AI-Powered WhatsApp Customer Support for Shopify Brands This n8n template builds a WhatsApp support copilot that answers order status and product availability from Shopify using LLM "agents," then replies to the customer in WhatsApp or routes to human support. ------------------------------------------------------------------------ Use cases - "Where is my order?" → live status + tracking link - "What are your best-selling T-shirts?" → in-stock sizes & variants - Greetings / small talk → welcome message - Anything unclear → handoff to support channel ------------------------------------------------------------------------ Good to know - WhatsApp Business conversations are billed by Meta/Twilio/Exotel; plan accordingly. - Shopify Admin API has rate limits (leaky bucket) --- stagger requests. - LLM usage incurs token costs; cap max tokens and enable caching where possible. - Avoid sending PII to the model; only pass minimal order/product fields. ------------------------------------------------------------------------ How it works 1. WhatsApp Trigger\ Receives an incoming message (e.g., "Where is my order?"). 2. Get Customer from Shopify → Customer Details → Normalize Input\ Looks up the customer by phone, formats the query (lower-case, emoji & punctuation normalization). 3. Switch (intent router)\ Classifies into welcome, orderStatusQuery, productQuery, or supportQuery. 4. Welcome path\ Welcome message → polite greeting → (noop placeholder). 5. Order status path (Orders Agent) - Orders Agent (LLM + Memory) interprets the user request and extracts needed fields. - Get Customer Orders (HTTP to Shopify) fetches the user's latest order(s). - Structured Output Parser cleans the agent's output into a strict schema. - Send Order Status (WhatsApp message) returns status, ETA, and tracking link. 6. Products path (Products Agent) - Products Agent (LLM + Memory) turns the ask into a product query. - Get Products from Shopify (HTTP) pulls best sellers / inventory & sizes. - Structured Output Parser formats name, price, sizes, stock. - Send Products message (WhatsApp) sends a tidy, human-readable reply 7. Support path Send a message to support posts the transcript/context to your agent/helpdesk channel and informs the user a human will respond ------------------------------------------------------------------------ How to use - Replace the manual/WhatsApp trigger with your live WhatsApp number/webhook. - Set env vars/credentials: Shopify domain + Admin API token, WhatsApp provider keys, LLM key (OpenAI/OpenRouter), and (optionally) your support channel webhook. - Edit message templates for tone, add your brand name, and localize if needed. - Test with samples: "Where is my order?", "Show best sellers", "Hi". ------------------------------------------------------------------------ Requirements - WhatsApp Business API (Meta/Twilio/Exotel) - Shopify store + Admin API access - LLM provider (OpenAI/OpenRouter etc.) - Slack webhook for human handoff ------------------------------------------------------------------------ Prerequisites - Active WhatsApp Business Account connected via API provider (Meta, Twilio, or Exotel). - Shopify Admin API credentials (API key, secret, store domain). - Slack OAuth app or webhook for human support escalation. - API key for your LLM provider (OpenAI, OpenRouter, etc.). ------------------------------------------------------------------------ Customising this workflow - Add intents: returns/exchanges, COD confirmation, address changes. - Enrich product replies with images, price ranges, and "Buy" deep links. - Add multilingual support by detecting locale and templating responses. - Log all interactions to a DB/Sheet for analytics and quality review. - Guardrails: confidence thresholds → fallback to support; redact PII; retry on API errors.