Why We Built a Custom AI Assistant
Off-the-shelf AI chatbots are generic. They do not understand your brand, your services, or your customers. Soedja built a custom AI assistant using Alibaba's Qwen model to deliver an experience that is deeply integrated with our service offerings and client needs.
The Technology Stack
Our AI assistant runs on Qwen, Alibaba Cloud's large language model, integrated into a Bubble.io frontend. This combination gives us the flexibility of a powerful language model with the rapid iteration speed of a no-code platform.
- Qwen LLM: Handles natural language understanding and generation.
- Custom system prompts: Trained on Soedja's services, pricing, and processes.
- Bubble.io frontend: Clean chat interface that matches our brand design.
What It Does
The assistant helps potential clients understand our services, get pricing estimates, and even begin scoping their project — all through a natural conversation. It can qualify leads, answer FAQs, and schedule consultations without human intervention.
Lessons Learned
Building a custom AI assistant taught us that the quality of the system prompt matters more than the model size. A well-crafted prompt with clear boundaries and contextual knowledge produces better results than a generic setup on a more powerful model.