One definition of every number. For people and for AI.
The semantic layer gives your business a single, governed vocabulary: measures, KPIs, hierarchies and synonyms defined once and used by every dashboard, report, alert and copilot answer.
Semantic model · Motor portfolio · v12
PublishedDraft v13ValidatePublish| Gross written premium | SUM(premium) · ₹ Cr |
| Claims ratio | claims_paid / earned_premium |
| Policies in force | COUNT DISTINCT policy_no · active |
| Avg. settlement days | AVG(settled_on − intimated_on) |
| Dashboards | 14 |
| Reports | 23 |
| Alerts | 6 |
| Ask Marut | all answers |
Three tools, three definitions of revenue, and an AI assistant that invents a fourth. Nobody can say which number is right, so nobody trusts any of them.
What makes it powerful
Business model, not tables
Measures, KPIs, hierarchies and relationships defined in business terms, over your pipelines and lakehouse.
Glossary and synonyms
“GWP”, “premium” and “topline” all mean the same thing, so people and the copilot can ask naturally.
Security in the model
Row-level rules live with the definitions and apply to dashboards, exports, APIs and AI answers alike.
Draft, validate, publish, roll back
Change a definition once, see what it affects, publish with a version, and roll back if needed.
Fast at scale
Columnar engine, cache and pre-aggregation behind every definition, so answers stay quick as data grows.
One foundation, every interface
Dashboards, reports, alerts, Ask Marut, voice, embedded views and the API all read the same model.
Business context for AI.
Marut AI does not read your tables. It reads your semantic model: what the measures mean, how they roll up, what people call them and who may see them. That is why its answers are right, cited and cheap to produce.
- Questions resolved against governed definitions
- Smaller prompts, fewer tokens, lower cost per answer
- Synonyms let people ask the way they talk
- The same permissions for AI as for everything else
Before and after
Without a semantic layer
- Every dashboard and report re-defines “revenue” its own way
- AI chatbots guess at raw tables and get it confidently wrong
- A rule change means editing dozens of reports
- Security is re-implemented per tool, and leaks through exports
With the Marut semantic layer
- One definition, used by every screen, file and answer
- The copilot reasons over your business terms, and cites them
- Change once, published with a version, everywhere at once
- Row and column rules enforced in the model, including for AI
What it does for your business
Numbers that match everywhere
The board pack, the dashboard and the AI answer agree, because they share one source.
AI you can trust
Grounded in definitions rather than raw columns, the copilot makes fewer mistakes and needs far less context per question.
Change once
A new KPI or a renamed hierarchy reaches every report and answer the moment it is published.
See your own KPIs modelled in 2 to 3 days.
Bring the definitions your teams argue about. We will model them once and show every screen and answer agreeing.