USE CASES / DATA GOVERNANCE AND TRUST
Data governance and trust: every number traces back to its definition.
AtScale versions and tests every definition, and one set of access rules covers every person, agent and tool.
People trust a number when they can see where it came from. AtScale manages your business semantics like code, with versions, reviews and tests, and passes each user’s identity through to the data platform. When an auditor asks about a number, you’ll be able to show the definition version and the SQL behind it.
The challenge
Definitions get edited in place with no review, test or rollback.
Access rules differ from one tool to the next.
You can’t tell which definition an AI answer used.
What you get
Every number is explainable to an auditor.
Access rules live in one place and follow every person and agent.
THE CHALLENGE
AI makes a wrong number more dangerous.
A dashboard with a wrong number gets questioned in a meeting. An agent with a wrong number acts on it, and an AI assistant repeats it to anyone who asks. When definitions live in BI tools, spreadsheets and prompts, there’s no single place to review a change. Nothing tests it before it ships, and you can’t see what an automated decision was based on.
HOW IT WORKS
Your business semantics, managed like code.
| What AtScale manages | How it works |
|---|---|
| Version history | Every definition lives in Git, with an author, a timestamp and a diff for each change. |
| Review and approval | Each change opens a pull request with a named reviewer, and only an approved version reaches production. |
| Regression testing | Canary reruns your test questions before and after each change and when you switch AI models. It’s coming this fall. |
| Lineage | Each metric points back to its source table and column, and every query traces to the definition version that produced it. |
| Execution record | Every live and test interaction is captured, and entries can’t be edited once they’re written. |
| Access control | Row- and column-level security is defined once, each user’s identity passes through to the data platform, and agents get their own identities. |
IDENTITY
One set of rules for people and agents.
AtScale signs people in through Okta, Microsoft Entra ID or Active Directory and passes their identity through to your data platform. An agent working for a regional analyst sees that analyst’s regions and can’t see any others, whether the question comes from Claude, Excel or Power BI.
THE PROOF
Built for regulated, audited businesses.
Blue Yonder manages 600 semantic objects in AtScale, with versions, reviews and permissions that apply in every tool. AtScale runs in production at large retailers, telecoms and financial services firms, inside their own environments, and their data doesn’t leave their data platform.
FAQ
Questions about data governance and trust.
Where do the definitions live?
In SML files in your own Git repository, one file per object.
Can we roll back a bad change?
Yes. Revert the pull request, and you’re back on the previous version in production.
Can we test a change before it ships?
Yes. Every change goes through a pull request with a named reviewer. Canary, coming this fall, will rerun your test questions before and after each change and when you switch AI models.
Does AtScale send our data anywhere?
No. ACE, the AI Computation Engine, computes each metric in your data platform, and results go straight to the tool that requested them.
Which identity providers does AtScale support?
Okta, Microsoft Entra ID and Active Directory (Kerberos). AtScale passes each user’s identity through to your data platform.
Do agents get more access than their users?
No. Each agent signs in with its own identity, and your access rules apply to it the same way they apply to a person.
Book a workshop.
The workshop takes thirty days from kickoff to results.
- Bring us your most controversial metric and the business questions your team and your agents ask about it every week.
- We benchmark your AI on them, with AtScale and without it.
- You keep the semantic model we built to run the test.