Shadow AI discovery and monitoring

Find the shadow AI in your company: which AI tools staff use, with which accounts, and what data goes into them.

Most companies have no list of the AI apps their people use. We use Microsoft Defender for Cloud Apps and Purview to discover AI tools across your devices, show which are used with personal accounts and what sensitive data reaches them, then help you decide what to allow, monitor or block.

Book an AI Data Risk Check
Layered AI workflow modules made visible along a continuous path.
The initiative

Move from a difficult technology challenge to a system your teams can use.

For IT and security leads who need to know how AI is really used before they write a policy or buy a tool. We begin with the business context and the systems already in place, then shape a delivery path around what needs to work.

  • AI app discovery across managed devices
  • Personal vs work account usage
  • Sensitive data shared with AI tools
  • Sanctioned and unsanctioned AI app lists
  • Ongoing AI usage monitoring and alerts
Explore relevant engineering stories
When this solution fits

Start with the work your team needs to make possible.

Shadow AI is any AI tool used for work without IT knowing: a free chatbot, a browser extension that summarises pages, a meeting note-taker, an AI feature switched on inside another app. You cannot set a sensible AI policy until you know what is in use. Discovery gives you the facts in days, using Microsoft tools many companies already license.

For IT and security leads who need to know how AI is really used before they write a policy or buy a tool. The first useful step is to define what needs to change, how the current environment shapes the work and which outcome the team needs to be able to assess.

Typical workstreams

What the engineering may involve.

The exact scope depends on your systems and constraints. These workstreams describe common parts of this solution area.

01

Discover

Turn on AI app discovery with Defender for Cloud Apps and Purview, and collect one to two weeks of real usage across managed devices.

02

Assess

Rank AI apps by risk, usage and data shared; separate work from personal accounts; flag the tools handling sensitive data.

03

Decide and monitor

Agree which tools to allow, monitor or block, apply the controls, and keep watching for new AI apps every month.

What shadow AI looks like

It is not one chatbot. It is a long list of AI tools nobody approved.

When we look at real usage, the list is usually longer than expected and spread across departments.

01

Public chatbots

ChatGPT, Gemini, Claude, DeepSeek, Perplexity and others, often on free personal accounts.

02

Browser extensions

AI writing, summarising and translation extensions that can read the pages staff open.

03

Meeting note-takers

Bots that join calls, record them and store transcripts outside your tenant.

04

AI inside SaaS apps

AI features switched on in CRM, design, HR and file-sharing tools without a review.

05

Code assistants

AI coding tools used on company source code, sometimes with personal accounts.

06

Home-made agents

Copilot Studio and other agents built by staff, connected to company data with no owner on record.

How discovery works

Real usage data in one to two weeks.

We use Microsoft tools many companies already license. Nothing is installed that slows people down, and no one’s conversations are read as part of discovery.

01

Turn on discovery

Defender for Cloud Apps discovery from Defender for Endpoint, plus Purview data security posture for AI where licensed.

02

Collect usage

One to two weeks of real activity across managed devices: which AI apps, how often, by which teams.

03

Flag the risk

Each AI app is scored for security and compliance; sensitive data sent to AI tools is flagged by type, not shown in full.

04

Report

A short report: AI apps by risk, personal vs work accounts, sensitive data trends and the decisions to make.

After discovery

Allow, monitor or block, then keep watching.

Discovery is the start. The value comes from acting on it and keeping the list current as new AI tools appear every month.

Self-check first: the free Microsoft 365 and AI security checklist

01

Sanctioned list

Approved AI tools, with work-account sign-in where the tool supports it.

02

Monitored list

Tools allowed for now, with alerts on sensitive data and heavy use.

03

Blocked list

High-risk or unnecessary AI apps blocked on managed devices with a clear message to users.

04

Monthly watch

New AI apps, usage changes and incidents reviewed every month.

Checklist

Signs you have a shadow AI problem

Several of these usually means AI use has moved ahead of your controls.

Request an assessment
A clear path from scope to operation

Keep decisions connected through delivery.

Move from a defined initiative to an operating system through focused, reviewable stages.

01

Understand the workflow

Agree on the users, business context, systems and constraints that define the initiative.

02

Shape the technical path

Connect architecture, integrations, data, security and delivery decisions to the needs already identified.

03

Build in useful increments

Deliver and review the system in stages so the work can respond to product and operational feedback.

04

Prepare to operate

Consider ownership, monitoring, support and the next improvements as part of the delivery plan.

Discuss this solution
FAQ

Questions about shadow ai discovery & ai usage monitoring

What is shadow AI?

Shadow AI is any AI tool used for work without the knowledge or approval of IT, such as a free chatbot on a personal account, an AI browser extension or an AI feature switched on inside another app.

How do you detect shadow AI?

With Microsoft Defender for Cloud Apps discovery, fed by Defender for Endpoint on managed devices, and Purview data security posture for AI. Together they show which AI apps are used, by how many people and what kinds of sensitive data reach them.

Do you read employees’ AI conversations?

Not as part of discovery. Reports show apps, usage and sensitive data types. Any deeper review follows your own policy and local law.

How long does it take?

Usually one to two weeks of data collection, then a report and a walkthrough call.

What licence do we need?

Defender for Endpoint P2 or Microsoft 365 E5 gives the most complete discovery. With other plans we use what is available and tell you what is missing.

What happens after discovery?

You decide which AI tools to allow, monitor or block. We apply the controls and can monitor new AI apps monthly.