About ChatOrder

The Problem We Saw

Every day, thousands of food businesses across the UK hand over 15–35% of their revenue to delivery platforms. A £25 order becomes £16 after commissions, fees, and VAT. For businesses running on 3–5% margins, this is the difference between surviving and closing.

Meanwhile, 71% of the UK population already has WhatsApp installed. Customers want to order directly. Businesses want to keep their revenue. The technology to connect them at production scale, with the safety and reliability of a real ordering platform, simply did not exist. ChatOrder is built to close that gap.

What ChatOrder Does

ChatOrder turns WhatsApp into a complete ordering channel. Customers scan a QR code or message a number, then chat with an AI assistant that knows the restaurant's menu, hours, delivery zones, and allergens. The assistant takes the order, validates the delivery address against the UK postcode service, calculates the bill, and sends a Stripe payment link inside the same chat. Funds settle directly to the restaurant's bank account. There is no app to install, no third-party marketplace, and no commission.

Innovation & Technology

ChatOrder is more than a chatbot wrapper. It is a multi-tenant, AI-native ordering platform that treats WhatsApp as a first-class commerce surface. The core technical contributions of the project include:

• Autonomous agent architecture. A layered agent (planner → executor → guardrails) runs every customer conversation. It selects from a catalogue of typed tools — add_to_cart, validate_address, checkout, confirm_order, lookup_order, modify_order, apply_promo_code, send_food_image — and is constrained by explicit guardrails that prevent hallucinated menu items, AI-identity leaks, and off-topic drift.
• Per-restaurant RAG.Each restaurant's knowledge base — opening hours, allergen policy, delivery zones, FAQs — is embedded with text-embedding-3-small and stored in pgvector. The agent retrieves only the entries relevant to the customer's question, keeping prompts small and answers grounded.
• Stateful customer memory. Saved addresses, preferred order type, payment preference, and favourite items persist across sessions. On a cache miss the system rehydrates from the canonical Django record so a flush or restart never loses customer context.
• Multi-tenant, multi-language, multi-currency. Restaurants in different countries run on the same instance, each with its own brand colour, currency (GBP, USD, EUR, PKR, INR, AED, SAR), opening hours, delivery zone logic, payment methods, and AI provider (OpenAI, Anthropic, Google, Groq).

The stack is open, standards-based, and built on production-grade components: Django + DRF + PostgreSQL + pgvector on the API tier, FastAPI + ARQ + Valkey on the messaging tier, Next.js on the dashboard, and the Meta WhatsApp Cloud API and Stripe Connect as external dependencies. Containerised, observable, and engineered to be cheap to operate.

Traction

ChatOrder is live with paying UK restaurants today. Live platform metrics — orders processed, customers served, businesses onboarded — are published on our live traction page and update directly from production. Pricing is pay-as-you-go at 99p per order; there is no monthly fee, no contract, and no commission.

About Devzenix Ltd

ChatOrder is built by Devzenix Ltd, a UK-registered technology company. We build infrastructure that helps small and medium businesses keep their margins by removing the platform middleman. We believe the best software disappears into the channels people already use.

Contact