AxonBuild - AI & Full Stack Solutions

Introduction

We built a secure and scalable no code solution that enables non technical users to deploy Retrieval Augmented Generation (RAG) chatbots for sales and customer support. Using Make.com as the automation backbone, the system integrates WhatsApp and webhooks with OpenAI embeddings and a Qdrant vector database. This approach allows businesses to set up AI powered sales assistants without writing a single line of code.

Challenge

Many small and mid sized businesses struggle to adopt AI chatbots because custom development is expensive, time consuming, and requires technical expertise. Traditional chatbots are either rigid decision trees or require complex coding for advanced retrieval capabilities. Businesses needed a solution that was simple to configure, secure, and capable of handling product FAQs, lead collection, and real time customer engagement at scale.

Solution

We designed a no code framework using Make.com's workflow builder that connects communication channels, AI models, and vector databases into a seamless pipeline.

No Code Data Upload

Using Make.com webhooks, we created a simple HTML page where users can upload product JSON files. Uploaded data is parsed, embedded, and stored in the Qdrant database automatically. This allows non technical staff to update product catalogs and knowledge bases without engineering support.

Scalable Automation

By leveraging Make.com modules and webhooks, the solution can be extended to other services such as email, CRM systems, or ticketing platforms without rewriting code.

WhatsApp Integration

A webhook monitors incoming customer texts on WhatsApp. Incoming messages are automatically passed to the workflow.

AI Response Generation

The retrieved information is passed to an AI agent which formulates an accurate and context aware response. This response is then delivered back to the customer through WhatsApp.

Embedding and Retrieval

Customer messages are embedded using an OpenAI API call. The embedding vector is then searched against a Qdrant collection that stores product information and metadata. The most relevant results are returned with similarity scores.

Technical Approach

No Code Automation:Make.com workflow builder for visual automation design

RAG Implementation:OpenAI embeddings with Qdrant vector database for retrieval

Messaging Integration:WhatsApp webhook monitoring and response automation

Data Management:Automated JSON parsing and embedding pipeline

Scalable Architecture:Extensible workflow design supporting multiple channels

Stack:Make.com scenarios, OpenAI APIs, Qdrant database, webhook integrations

What we've accomplished

Successfully delivered a no code RAG chatbot platform that democratizes AI deployment for small and mid sized businesses. The solution eliminates technical barriers while maintaining sophisticated retrieval capabilities and scalable automation workflows.

Results & Impact

•

Setup Time:Reduced from weeks of coding to hours of drag and drop configuration

•

Ease of Use:Non technical users can add or update product data directly via a web page

•

Cost Savings:Eliminated the need for custom chatbot development and maintenance overhead

•

Scalability:Secure and extensible architecture that supports additional channels and workflows

•

Adoption:Enabled rapid deployment for sales teams with minimal training

Project Narrative

The journey began with a need for a simple sales assistant that could answer product FAQs and engage customers on WhatsApp. Using Make.com scenarios, we connected messaging, embeddings, and retrieval into a fully functional RAG chatbot with no custom code. Next, we extended the workflow to allow data uploads via a simple web page, so that staff could refresh product catalogs without engineering help. Finally, by embedding new data into a Qdrant collection, the chatbot became continuously up to date. What started as a manual sales query process evolved into a secure, no code AI platform that empowers non technical teams to run sophisticated RAG chatbots at scale.