---
title: "How to ground Genie Agents in both structured data and documents without losing governance"
description: "Databricks Genie Agents can combine structured tables and unstructured documents, but the central challenge is preserving data governance when one agent can access both. The proposed architecture runs agents with the end user’s credentials, using Unity Catalog identity synchronization, object privileges, ABAC policies, row filters, and column masks to enforce permissions at query time. Automatic Identity Management syncs users and groups from Microsoft Entra ID and Okta, while just-in-time provisioning carries current memberships into Databricks. Structured assets include tables and metric views, and governed tags can attach protection across newly tagged data; documents are governed through Unity Catalog volumes and their grants. In the Brickstore example, APAC and AMER managers ask the same question yet receive different, authorized results, with regional rows, masked email values, and shared metric definitions; the source recommends testing by impersonating users and planning volume boundaries."
---

# How to ground Genie Agents in both structured data and documents without losing governance

[Databricks](https://yomu.fyi/company/databricks) · Doyoung Jung · Aug 10, 2026

**Type:** Explainer

## Summary

Databricks Genie Agents can combine structured tables and unstructured documents, but the central challenge is preserving data governance when one agent can access both. The proposed architecture runs agents with the end user’s credentials, using Unity Catalog identity synchronization, object privileges, ABAC policies, row filters, and column masks to enforce permissions at query time. Automatic Identity Management syncs users and groups from Microsoft Entra ID and Okta, while just-in-time provisioning carries current memberships into Databricks. Structured assets include tables and metric views, and governed tags can attach protection across newly tagged data; documents are governed through Unity Catalog volumes and their grants. In the Brickstore example, APAC and AMER managers ask the same question yet receive different, authorized results, with regional rows, masked email values, and shared metric definitions; the source recommends testing by impersonating users and planning volume boundaries.

## Context

A single agent may access both structured and unstructured business data, creating a governance risk if it returns information to users who are not authorized to see it. The source contrasts this with systems that rely on prompt instructions as the security boundary and describes the need for current enterprise identities and unified controls.

## Approach / What changed

Run Genie Agents with end-user credentials and enforce access through Unity Catalog. Identity Management and just-in-time provisioning keep memberships current; object privileges, ABAC policies with governed tags, row filters, and column masks control structured data, while Unity Catalog volumes and grants govern documents. Validate behavior by asking the same question as users in different groups.

## Takeaways

- Unity Catalog, rather than the language model, remains the security perimeter: data is filtered at query time before an answer leaves the Lakehouse.
- ABAC policies and governed tags let one protection rule propagate to newly tagged data, reducing repetitive table-by-table configuration.
- In the Brickstore example, APAC and AMER managers receive different authorized results from identical prompts, including region-specific rows and masked customer email values.

**Tags:** [Architecture](https://yomu.fyi/topic/architecture), [Databricks](https://yomu.fyi/topic/databricks), [Privacy](https://yomu.fyi/topic/privacy)

- Source: [Databricks](https://www.databricks.com/blog/how-ground-genie-agents-both-structured-data-and-documents-without-losing-governance)
- Source URL: https://www.databricks.com/blog/how-ground-genie-agents-both-structured-data-and-documents-without-losing-governance
- Ingested by Yomu: 2026-08-30T16:51:46.604Z

[Read original post](https://www.databricks.com/blog/how-ground-genie-agents-both-structured-data-and-documents-without-losing-governance)