Fortune-level design-build enterprise
Self-Service Analytics Agent
Packaged a governed semantic layer into a natural-language analytics tool. Non-technical users ask questions in plain English and get answers routed to the right domain expert, with no SQL required.
- Headline number
- 11 domain-expert skills
- Role
- Built
- Category
- AI Agents & Runtimes
- Updated
Architecture
- Business userplain English
- Intent router
- 11 domain skillsHR, projects, precon, more
- Cortex Analystgoverned SQL
- Semantic views
Key takeaways
- 11 domain-intelligence skills, each an expert on one slice of the business
- Intent routing picks the right skill, so users never choose a dataset
- Every answer runs through the governed semantic layer, so numbers match finance
- Zero SQL for the user, full SQL visible for the auditor
The situation
The semantic layer was live and trusted, but using it still meant a BI tool or a SQL console. Department leads wanted to ask a question and get an answer during a meeting, with the same rigor the analysts had.
What I built
A Claude Code plugin that packages the semantic layer as a set of skills. Each skill knows one domain deeply: which semantic view to use, which metrics are safe to combine, what the common questions are, and what caveats to attach. A router reads the question, picks the skill, and the skill generates a governed query through Cortex Analyst.
The consumption layer sits on the same 15 views as the rest of the stack. I counted it once. This is a product on top of the data layer, never a second copy of it.
The outcome
People who had never opened a query tool answer their own capacity and pipeline questions. The analysts stopped fielding the repetitive ones and spend their time on the hard ones. Every answer carries its query, so trust went up rather than down.
Stack
Related work
Same discipline, different system.
Need this built inside your organization?
One person, architecture to deployment. Tell me what is stuck and I will tell you where I would start.