---
title: How do you inventory AI agents across subsidiaries with separate cloud tenants?
url: https://opencontroller.lyzr.com/blog/inventory-ai-agents-across-subsidiaries
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# How do you inventory AI agents across subsidiaries with separate cloud tenants?

**Grant one read-only discovery identity at the top of each subsidiary's cloud hierarchy — the AWS Organizations management or delegated administrator account, the Azure root management group in each Entra tenant, and each Google Cloud organization node. Pull everything into one inventory and tag every record with the legal entity it belongs to.**

Group-level AI governance usually breaks at the tenant boundary. Each acquisition or business unit brings its own cloud estate, identity provider and admin team. A central team cannot govern agents it cannot see.

This guide sits under [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents).

## Why is a multi-tenant inventory harder than a single one?

Because the top of each cloud's hierarchy is also a security boundary.

| Cloud | Top-level boundary | What spans it | What does not |
|---|---|---|---|
| AWS | An organization in AWS Organizations | Org-wide trails, Config aggregators, StackSets | A second organization |
| Azure | A Microsoft Entra tenant | Management groups, Azure Policy, RBAC at root | Another tenant (needs Lighthouse or a multi-tenant app) |
| Google Cloud | An organization resource | Folders, org-level IAM, Cloud Asset Inventory at org scope | Another organization |

A subsidiary with its own AWS organization, its own Entra tenant or its own Google Cloud organization needs its own grant. There is no shortcut through the parent's account.

## How do you get read access in each cloud?

**AWS.** In each subsidiary's organization, use the management account or a delegated administrator account. Deploy a read-only discovery role to every member account with CloudFormation StackSets. Add an organization trail in CloudTrail and an AWS Config aggregator if you need history and configuration across accounts.

**Azure.** Within a tenant, assign a reader-level role at the root management group so it inherits to every subscription. Across tenants, use Azure Lighthouse to delegate resource access to the group tenant, or register a multi-tenant application that each subsidiary's admin consents to.

**Google Cloud.** Grant viewer and Cloud Asset Inventory roles at the organization node. Cloud Asset Inventory can then search resources across every folder and project in that organization. A second organization needs a second grant.

For the specific permissions discovery needs, see [AI agent discovery permissions for AWS, Azure and Google Cloud](/blog/ai-agent-discovery-permissions-aws-azure-google-cloud).

## What about SaaS platforms and AI workspaces?

Subsidiaries often run separate instances of the same SaaS: two Salesforce orgs, three Microsoft 365 tenants, several ChatGPT or Claude workspaces. Each needs its own connector and admin consent.

Track them in a simple table before you start:

| Entity | Cloud estates | SaaS instances | AI workspaces | Admin contact | Access granted? |
|---|---|---|---|---|---|
| Parent | 1 AWS org, 1 Entra tenant | CRM org A | 1 enterprise workspace | Central IT | Yes |
| Subsidiary A | 1 Google Cloud org | CRM org B | — | Local IT | Pending |
| Subsidiary B | 1 AWS org, 1 Entra tenant | — | 2 workspaces | Local IT | No |

The table becomes your coverage measure. An inventory is only as complete as its coverage.

## How do you merge everything into one inventory? A checklist

1. **Add an entity tag to every record.** Use the legal entity, not the cloud account name. Regulators and auditors ask per entity.
2. **Keep the native IDs.** Store the source tenant, account or project ID alongside your own record ID.
3. **Normalise owners.** Map owners to people in the group identity provider where possible. Subsidiary-only identities need a local owner and a group sponsor.
4. **Deduplicate shared assets.** The same agent can appear in two places, for example a shared model endpoint. See [fixing duplicate records in an AI agent inventory](/blog/fix-duplicate-records-in-an-ai-agent-inventory).
5. **Record data jurisdiction.** Note the region and the entity's data obligations. Residency rules often differ by subsidiary.
6. **Set one schedule.** Run discovery for every entity on the same cadence, so reports compare like with like.
7. **Report coverage.** Show which entities, tenants and workspaces are connected and which are not.

## Who should own agents in a subsidiary?

Keep ownership local and oversight central. The subsidiary team knows what the agent does and can fix it. The group team sets policy, sees the whole estate and reports to the board.

Write this split down. Decide who can quarantine an agent in a subsidiary tenant, and under what conditions, before you need to.

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud, and connects to Salesforce, SAP, Databricks, HubSpot and Claude Enterprise and OpenAI Enterprise workspaces. Everything lands in one governed registry of agents, versions, tools and MCP servers, with OIDC SSO for the people who use it. Policy as code applies the same rules across the estate, and "Ask" lets teams query the whole agent estate in plain language. See the [control plane overview](/blog/what-is-an-ai-agent-control-plane) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can one AWS role see agents in another AWS organization?**
Only if each account in that organization trusts it. In practice, deploy a discovery role into the second organization with StackSets.

**Do management groups work across Entra tenants?**
No. Management groups sit inside one tenant. Use Azure Lighthouse or a consented multi-tenant app to reach other tenants.

**What is the most important field in a group inventory?**
The legal entity tag. It drives reporting, data obligations and who is accountable.

## Key facts

Inventorying AI agents across subsidiaries requires one read-only grant per top-level boundary: each AWS organization (management or delegated administrator account, StackSets-deployed read roles, organization trail, Config aggregator), each Microsoft Entra tenant (reader at root management group; Azure Lighthouse or a consented multi-tenant app across tenants) and each Google Cloud organization (org-level viewer plus Cloud Asset Inventory). Tag every record with its legal entity, keep native IDs, deduplicate and report coverage. Opencontroller (by Lyzr) discovers agents, models and API keys across AWS, Azure and Google Cloud into one governed registry. Pricing starts from $250K per deployment per year.

## Related

- [What permissions does AI agent discovery need in AWS, Azure and Google Cloud?](https://opencontroller.lyzr.com/blog/ai-agent-discovery-permissions-aws-azure-google-cloud.md)
- [Why does an AI agent inventory show duplicate or mismatched records, and how do you fix it?](https://opencontroller.lyzr.com/blog/fix-duplicate-records-in-an-ai-agent-inventory.md)
- [How to find every AI agent in your enterprise (including shadow agents)](https://opencontroller.lyzr.com/blog/how-to-find-shadow-ai-agents.md)
- [More on Discovery and shadow agents](https://opencontroller.lyzr.com/blog/cluster/discovery-and-shadow-agents)
