Chat with governed Looker models
Ask questions in plain language over LookML-defined explores and get answers that use the company's agreed metric logic.
The use case
Business users want to ask questions without learning Looker's explore interface, but the company wants answers to use agreed definitions. Conversational Analytics in Looker answers natural-language questions using LookML explores, so metrics and joins come from the semantic layer. Data teams can build agents scoped to particular explores with extra instructions. Answers show which fields and filters were used. Analysts maintain the LookML, review questions that fail, and users confirm surprising answers before acting on them. A small pilot group on one explore shows quickly where field names and descriptions need work.
Example workflow
- Choose explores to expose
- Add instructions and field descriptions
- Users ask questions in chat
- Review the fields and filters used
- Improve LookML from failed questions
Tools
- Looker
- Gemini
Risks and guardrails
- Missing field descriptions lead to wrong field choices
- Permissions follow Looker roles; check them
- Do not use unreviewed answers in financial reporting
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Google Cloud docs: Conversational Analytics in Looker: docs.cloud.google.com/looker/docs/conversational-analytics-overview
- Gemini (official page): gemini.google.com (checked 2026-11-09)