AI Supplier Order Management:

Automated Order Collection, Parsing & Dispatch for a Restaurant & Bar


🏒 Context

Company: Kitchen & Bar - full-service restaurant and bar venue Industry: HoReCa (Restaurants, Bars, Cafes) Target clients: Venue staff (bar, kitchen) + operations manager Order volume: Daily supplier orders across 2 departments, 4 suppliers Team: Bar staff + kitchen staff + 1 manager Core problem: Orders collected via WhatsApp and paper notes - no consolidation, no control, no audit trail


❌ Problem

Metric Value Issue
Order collection 100% manual WhatsApp messages, paper notes, verbal requests
Duplicate detection None Same item ordered twice from different departments
Supplier consolidation Manual Manager merged orders by hand before sending
Order confirmation None No visibility into what was sent and when
Audit trail None No log of what was ordered, by whom, for which supplier

The manager received orders from bar and kitchen staff through different channels with no structure. Consolidating orders per supplier before sending was done manually, took time, and regularly caused duplicates or missed items.


βœ… Solution

Built a 2-workflow AI order automation system that collects, parses, consolidates, and dispatches:

[bar.html / restaurant.html]
        β”‚
        β”‚  POST { venue: "bar"|"restaurant", items: "raw text" }
        β–Ό
[WF-A: Parse & Store]  ←── Webhook: /webhook/smartorder-demo
        β”‚
        β”œβ”€ Read Products Master (Supabase)
        β”œβ”€ AI Parse (Claude Haiku) β€” item matching + typo handling
        β”œβ”€ Map to Rows (Code node)
        └─ Insert β†’ orders_parsed (status: pending)
                        β”‚
[WF-B: Manager Bot]  ←── Telegram Trigger (manager messages bot)
        β”‚
        β”œβ”€ /digest or text  β†’  digest by supplier + inline buttons
        β”œβ”€ view_*           β†’  supplier detail view
        β”œβ”€ send_*           β†’  send to supplier β†’ status: sent
        β”œβ”€ send_all         β†’  dispatch to all suppliers at once
        └─ keep_open        β†’  stay pending, allow more additions

WF-A - Order Intake: Parse & Store

Triggered by staff submitting free-text orders via web terminal (bar.html or restaurant.html). Claude Haiku reads the raw input, matches each item to the products master table in Supabase, identifies the correct supplier, and saves each line item as a separate row with status: pending. Typos and informal naming (e.g.Β β€œhendrix” β†’ Hendricks Gin) handled automatically.

WF-B - Manager Bot: Review & Dispatch

Telegram bot interface for the manager. On any message or /digest, returns a consolidated view of all pending items grouped by supplier - Bar and Restaurant orders automatically merged. If both departments ordered the same item, it is flagged with πŸ”€. Manager taps a supplier to see the full item list, then approves and dispatches with one button.


βš™οΈ Process

1. AI Parsing Logic