How Airi uses your knowledge base

Your knowledge base powers grounded answers for agents and AI auto-reply. This page explains what Airi retrieves, how answers are built, and how to keep response quality high.

Why this matters

• Keeps support answers consistent across managers, channels, and shifts.
• Reduces guesswork by grounding replies in your own policies and docs.
• Improves AI coaching quality because suggestions use verified context.
• Shortens response time for repeated questions.

What Airi reads

Airi combines multiple knowledge sources before generating a reply or suggestion.

• Published Knowledge Base articles
• Verified Q/A snippets
• Indexed public documentation links
• Current conversation context and language

How retrieval works

1. Index your sources - When you publish KB content, Airi stores searchable representations so relevant sections can be found quickly.

2. Match incoming question - For each new customer message, Airi looks for the most relevant KB blocks and snippets.

3. Build grounded answer - Airi drafts a response based on retrieved context, prioritizing your KB over generic assumptions.

4. Coach and improve - Managers get guidance that references available knowledge, making quality control and onboarding easier.

Best practices for reliable replies

• Write short, specific articles around one topic each.
• Start with top recurring questions from real conversations.
• Update policies and pricing articles immediately after changes.
• Use clear article titles so retrieval relevance stays high.
• Review unanswered or weak replies weekly and fill KB gaps.

Tip: Start with 10 high-frequency customer questions first, then expand your KB every week.