---
title: How do you find AI features switched on inside SaaS apps?
url: https://opencontroller.lyzr.com/blog/find-ai-features-inside-saas-apps
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# How do you find AI features switched on inside SaaS apps?

**Combine three reviews: the admin console of each SaaS app (for AI settings and agent features), the vendor's contract and sub-processor list (for which model providers see your data), and OAuth grants in your identity provider (for AI tools that connected themselves to your data). Together they find embedded AI that network and cloud scans miss.**

Many SaaS vendors have added assistants, summarisers and agents to products you already pay for. These features often run inside the vendor's cloud, so they never appear in your own cloud accounts or as traffic to a model provider.

## Why don't normal discovery methods find embedded AI?

Cloud discovery finds agents and models in your AWS, Azure or Google Cloud accounts. Network and browser monitoring finds people visiting AI websites. Embedded SaaS AI does neither: your user clicks "summarise" inside a tool you approved years ago, and the vendor calls a model from its own infrastructure.

So the evidence lives in three other places: the app's settings, the vendor's paperwork and your identity provider.

## Step 1: Review each app's admin console

Start with the apps that hold the most sensitive data: email, documents, chat, CRM, HR, ticketing and finance.

For each app, record:

1. Whether AI features exist on your plan.
2. Whether they are on by default, off by default, or on for some users.
3. Whether admins can switch them off per workspace, group or user.
4. Whether the vendor offers agent features that can take actions, not just summarise.
5. Whether AI usage appears in the app's audit log.
6. Whether data used by AI features is retained, and for how long.

Many vendors put these controls in a dedicated AI or "generative features" section of the admin settings. If you cannot find one, ask the vendor in writing.

## Step 2: Review contracts and sub-processor lists

Vendors that send your data to a third-party model provider normally have to list that provider as a sub-processor. Check:

| Document | What to look for |
|---|---|
| Sub-processor list | New entries for model providers or AI infrastructure since your last review |
| Data processing agreement | Whether your data may be used to train models, and how to opt out |
| AI addendum or terms | Retention, region of processing, human review of inputs |
| Release notes | Announcements of new AI or agent features |
| Change notifications | Whether you subscribed to sub-processor change notices |

Subscribe to sub-processor change notices for your top 20 apps. It is the cheapest early warning you have.

## Step 3: Review OAuth grants in your identity provider

The third category is AI tools that connected themselves to your SaaS data: meeting note-takers, writing assistants, browser extensions and agent builders that asked users to "Sign in with" your identity provider and granted access to mail, calendar or files.

- In Microsoft Entra ID, review enterprise applications and the permissions users and admins consented to. Microsoft documents how to [review permissions granted to applications](https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-application-permissions).
- In Google Workspace, the Admin console lists third-party apps with access to your data under app access control. See Google's guide to [controlling which third-party apps access Workspace data](https://support.google.com/a/answer/7281227).

Sort grants by scope. Apps with read access to all mail or all files deserve review first. Restrict user consent so new AI apps need admin approval.

## How do you keep this current?

Embedded AI changes with every vendor release. Make it routine:

1. Re-run the admin console review quarterly for the top 20 apps.
2. Watch sub-processor notices continuously.
3. Review new OAuth grants weekly.
4. Add an "AI features" question to procurement and renewal checklists.
5. Record each embedded AI feature as an entry in your AI inventory, with an owner, so it gets the same review as agents you build.

For AI use that happens in browser tabs rather than inside apps, see [how to discover AI use in the browser](/blog/how-to-discover-ai-use-in-the-browser). For the wider method, start with the pillar on [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents).

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud, and has connectors to third-party agent platforms including Salesforce, SAP, Databricks and HubSpot, plus Claude Enterprise and OpenAI Enterprise workspaces. Agents found through those connectors land in a governed registry with owners and versions, alongside agents you build yourself, and per-agent Block and Quarantine controls act through the connected platform. For SaaS apps without a connector, you can register their AI features in the same registry. [Book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Are embedded SaaS AI features shadow AI?**
They are when nobody in security, privacy or IT has reviewed them. The app may be approved while the AI feature inside it is not.

**Should we turn embedded AI features off by default?**
For apps holding regulated or confidential data, switch them off until reviewed, then enable per group. For low-risk apps, review and document instead.

**What is the fastest first step?**
Export OAuth grants from your identity provider and sort by scope. It usually surfaces the most data exposure in the shortest time.

## Key facts

To find AI features inside SaaS apps, review each app's admin console for AI and agent settings, review vendor sub-processor lists, data processing agreements and AI terms for model providers and training use, and review OAuth grants in Microsoft Entra ID or Google Workspace for AI tools with access to mail, calendar or files. Repeat quarterly and record features in the AI inventory. Opencontroller by Lyzr discovers agents across AWS, Azure and Google Cloud and connects to Salesforce, SAP, Databricks, HubSpot, Claude Enterprise and OpenAI Enterprise workspaces. Pricing starts from $250K per deployment per year.

## Related

- [How do you discover employees' AI use in the browser?](https://opencontroller.lyzr.com/blog/how-to-discover-ai-use-in-the-browser.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)
- [How do you govern AI agents built by vendors and system integrators?](https://opencontroller.lyzr.com/blog/govern-ai-agents-built-by-vendors.md)
- [More on Discovery and shadow agents](https://opencontroller.lyzr.com/blog/cluster/discovery-and-shadow-agents)
