Multi-Store Order Allocation Workflow with OpenAI, Postgres & Slack

Automatically allocate multi-store inventory for every new order—then confirm or get ops approval via Slack

This n8n workflow receives new order webhooks, uses OpenAI to plan and execute inventory reservations across multiple stores via HTTP APIs, logs the full allocation decision trail to Postgres, and either auto-confirms the order in your OMS or sends a Slack review request for partial/failed fulfillment.

What this workflow does

• Ingest orders via Webhook: Accepts a POST webhook containing order ID, customer ID, line items, and optional candidate stores.
• Prevent duplicate processing: Creates an order-scoped idempotency key and checks Postgres to ignore deliveries already processed for the same order.
• Plan + reserve with OpenAI: Uses an OpenAI agent with HTTP tools to check per-store stock, respect rate limits, reserve inventory idempotently, and release reservations when a multi-store plan must be rolled back.
• Audit every decision: Parses the agent’s JSON output and writes the allocation plan, reservations, verdict, and any shortfalls to Postgres for traceability.
• Confirm or escalate: If the verdict is to confirm, sends the allocation plan to the OMS confirmation endpoint. If review is needed, posts a shortfall summary and allocation details to Slack, waits for an ops decision, then confirms or cancels via the OMS API.

Use cases

• High-volume e-commerce operations that need multi-store fulfillment decisions without manual spreadsheets.
• Teams that require reservation audit logs in Postgres for operational transparency.
• Order flows where partial stock must trigger a Slack approval gate before OMS confirmation.

Technical details

• n8n nodes: if, set, no op, wait, slack, webhook
• Integrations: OpenAI (agent), Postgres (e.g., tables like inventory_orders and inventory_reservations), Slack, HTTP tools for per-store stock/reservation and OMS endpoints
• OpenAI model: Uses gpt-4.1 (ensure it’s available in your OpenAI account)