Amazon Kendra
- What it is
- Intelligent enterprise search powered by machine learning. Users ask natural-language questions and get a specific answer, not a list of links.
- Key traits
- Connectors index content across silos: Amazon S3, SharePoint, Salesforce, ServiceNow, Confluence, RDS, and more.
- Returns three answer types: a direct factoid answer, a relevant FAQ match, and ranked document excerpts.
- Respects source-system access control lists, so users only see what they are permitted to see.
- Incremental learning from click feedback improves ranking over time.
- Not to be confused with
- Amazon OpenSearch Service — a general search and log analytics engine you tune with keyword relevance. Kendra is natural-language question answering out of the box, no relevance tuning.
- Amazon Q — generative AI assistant; Kendra is often its retrieval layer.
- Exam signal
- “Natural-language search across internal company documents” → Kendra.