Understanding how NovaSupport trains your AI agent helps you get better results and more accurate answers for your customers.
The Training Pipeline
When you add content to your Knowledge base, NovaSupport processes it through several stages:
- Ingestion: We read and parse your documents, websites, or text.
- Extraction: Our AI identifies key facts, questions, and answers.
- Verification: You review and approve the extracted information.
- Indexing: Verified facts are indexed for fast retrieval.
Why Q&A Extraction Matters
Unlike simple keyword matching, our Q&A extraction process understands context. When you upload a product manual, we don't just store the text – we understand that "The battery lasts 8 hours" is an answer to "How long does the battery last?"
Continuous Learning
Your agent gets smarter over time. When you:
- Add new knowledge sources
- Edit or approve suggested Q&As
- Review chat logs and fill knowledge gaps
Each action improves your agent's accuracy and usefulness.
The Verification Layer
Every piece of information your agent uses goes through your verification. This means:
- No hallucinations or made-up answers
- Complete control over what your agent knows
- Easy updates when information changes