Agree on what the number means. Then ask AI.

When “active customer” means something different in each conversation, a fluent answer can still be the wrong answer. Give people, agents and reports published business definitions to work from.

Business meaning belongs in the definition.

Define concepts and synonyms, metrics, dimensions, filters and joins. Review and publish the meaning your organization wants AI to use.

Keep suggestions and edits distinct from approved definitions. Saving a change does not automatically approve it for use.

“How many customers renewed?” needs a few agreed details.

Illustrative definition: count distinct customer accounts with a completed renewal in the reporting period. Specify the renewal event, account relationship, date field and exclusions before interpreting the number.

PrivacyFirst can compile governed semantic questions against supported sources, using published definitions and validated joins. Read-only execution, live permissions and provenance remain part of the path.

One definition can travel further than one answer.

Reuse governed business meaning in agent tools and semantic report widgets. Viewer permissions still determine what can be queried; sharing a report does not grant access to the underlying data.

Supported dbt and Open Semantic Interchange imports create drafts for review. Reuse existing work while checking the supported format and its mapping before publication.

Bring the metric that starts a debate.

Compare the intended business meaning with its sources, filters and joins. Test representative questions against a known expected result.

Review a proposed change, inspect its publication state, and repeat the query under another person’s permissions. Evaluate the explanation and provenance alongside the number.

Discover and understand source data

Build useful reports

Start with one metric your team needs to trust.

We can work through its definition, authorized query path and use in an actual business task.

Explore a semantic use case

Discuss your use case