How ConversAI uses AI to generate responses
To explain how ConversAI uses AI tools to generate responses, what content is used, and how user data is handled.
Target audience:
Administrators and technical users evaluating ConversAI’s AI architecture and data handling.
Overview
ConversAI uses Microsoft Azure’s AI services to generate responses based on customer-provided content. It does not train or fine-tune any AI models with internal data. Instead, it uses a Retrieve-Augment-Generate (RAG) architecture to deliver accurate, structured answers.
Key components
- AI tools used
- Azure AI Search
- Azure OpenAI Service (GPT-4o-mini)
- Content source
- Only customer-created content is used
- No external web or GPT-generated content is included
- Response generation
- ConversAI retrieves relevant content from the customer’s Mavim Azure Tenant
- GPT-4o-mini processes this content in inference mode
- The model generates a structured response based on the retrieved context
- No human intervention is required during response generation
- Model training
- ConversAI does not train or fine-tune any AI models
- GPT-4o-mini is used in inference mode only
- Customer data is never used to train the model
Important notes
- ConversAI operates in a closed ecosystem
- Prompts and outputs are not used to train the AI
