Inspiration
Many businesses sell through WhatsApp, Instagram, and direct messages, but managing customer conversations manually can quickly become overwhelming. Customers ask about availability, prices, delivery, services, bookings, and order status, often at the same time. If the business owner replies late or loses track of details, they can lose sales.
I built ChatOrder Autopilot Agent to solve that problem: an AI assistant that can talk to customers, understand what they want, help them place orders, and send the order details to the business dashboard automatically.
What it does
ChatOrder Autopilot Agent gives each business a public AI ordering link. Customers open the link, chat naturally, ask questions, browse available products or services, and place orders.
The system supports:
- Business-specific AI behavior
- Product and service catalogs
- Customer details collection
- Order creation and order status tracking
- Stock reduction for products
- Deposit settings and payment proof upload
- Business owner dashboard
- New order notifications inside the dashboard
- Qwen-powered AI conversation handling
How we built it
The project was built as a Node.js and Express web application with MongoDB for persistent storage. The frontend is served as static HTML, CSS, and JavaScript, with separate experiences for business owners and customers.
The AI layer uses Qwen through Alibaba Cloud/DashScope-compatible API settings. Customer messages are processed by the backend, where the system combines business profile data, product/service catalog data, customer conversation state, and order rules before generating a response.
The application was deployed on Alibaba Cloud ECS using Ubuntu, Node.js, PM2, and security group configuration to expose the web server publicly.
Challenges we ran into
One major challenge was making the AI conversational while still keeping order logic reliable. The assistant needed to understand natural messages like “I want two vanilla cakes,” “show my last order,” or “when should I expect it?” without breaking the order flow.
Another challenge was supporting different business types. The system had to work not only for products like cakes or water, but also for services like nails, bookings, packages, and deposits.
Deployment also came with real-world issues, such as configuring environment variables, running the app persistently with PM2, opening the correct Alibaba Cloud security group port, and fixing browser behavior differences between HTTPS Render deployment and HTTP Alibaba Cloud deployment.
Accomplishments that we're proud of
I’m proud that ChatOrder is no longer just a demo. It is a working AI-powered ordering system with a real customer chat page, business dashboard, Qwen AI integration, order management, payment proof upload, and Alibaba Cloud deployment.
The system can already be tested by real businesses and customers through a public link.
What we learned
I learned how to connect AI conversation with real backend actions, not just chatbot replies. This project helped me understand how to design conversation state, product matching, order confirmation, business-specific prompts, and deployment workflows.
I also learned how to deploy and operate a Node.js app on Alibaba Cloud ECS using PM2, environment variables, security groups, and Qwen API integration.
What's next for ChatOrder Autopilot Agent
Next, I want to improve ChatOrder with:
- WhatsApp/email notifications for business owners
- Automated payment confirmation through providers like Paystack
- Custom domains and HTTPS for trusted business links
- Better analytics for orders and customer behavior
- More advanced AI memory and business-specific training
- A marketplace/discovery page where customers can find businesses
The long-term goal is to make ChatOrder a reliable AI business assistant that helps businesses respond faster, sell more, and manage customer orders with less manual work.
Built With
- alibaba-cloud-dashscope-compatible-api
- alibaba-cloud-ecs
- css
- express.js
- github
- html
- javascript
- mongodb
- mongoose
- node.js
- pm2
- qwen-ai
- render
- ubuntu
- vanilla-javascript
Log in or sign up for Devpost to join the conversation.