How do I find out which AI tools employees are already using?

By Hugh Donatello, LAN Service Group · Published October 25, 2026 · Updated October 25, 2026

✓ Checked against Microsoft Learn, Google Workspace Help and Chrome Enterprise Help documentation on October 21, 2026.

Short answer

Look in five places: your identity provider's sign-in and app-consent records (Microsoft Entra ID or Google Workspace), a cloud-app discovery tool such as Microsoft Defender for Cloud Apps filtered to its Generative AI category, the browser extensions on managed devices, expense reports and card statements, and the people themselves. Each source sees a different slice and none sees everything, so combine them, then sort every tool you find into approve, approve with limits, or retire in an approved-tools register.

Free AI acceptable-use policy template and approved-tools register

Editable Word document and PDF, plus an Excel register of approved AI tools. No form or sign-up. Fill in the bracketed fields and have your own legal and HR advisers review the policy before you adopt it. What the template covers →

Download Word (.docx) Download PDF Download approved-tools register (.xlsx)

Why one source is never enough

AI tools reach a company in different ways, and each leaves a different trace. A tool someone signs in to with their work Microsoft or Google account shows up in identity logs. A tool that asks to read mail, files or calendar leaves a consent record. A chat site used in a browser with a personal account leaves only network traffic. A browser extension leaves an entry in the browser. A paid subscription on a personal card leaves an expense claim, if anyone claims it. Discovery means checking each of those trails, knowing what each one misses, and writing the results into one list.

This page is about finding what is already in use. Deciding what to do with a specific tool is a separate step: our guide to vetting an AI tool covers the review, and our guide to blocking or allow-listing AI tools covers the controls. If you are starting from scratch, read what shadow AI is and why it matters first.

Microsoft 365: Entra sign-in logs and app consent

Start with the Microsoft Entra sign-in logs. Microsoft says Entra logs all sign-ins into a tenant and that each entry shows who signed in, the client application used and the resource accessed. To view them, sign in to the Microsoft Entra admin center as at least a Reports Reader and go to Entra ID > Monitoring & health > Sign-in logs, then filter by application. Any AI tool that uses 'Sign in with Microsoft' on a work account appears here. Tools used with a personal account, or with a separate username and password, do not.

Check retention before you rely on history. Microsoft's data retention table lists seven days of sign-in and audit logs for Microsoft Entra ID Free and 30 days for P1 and P2, and says upgrading does not bring back data that has already expired. If you need a longer view, Microsoft says you can route the logs to an Azure storage account through Azure Monitor.

Next, look at what users have consented to. Microsoft records a 'Consent to application' activity in the Entra audit logs (service Core Directory, category ApplicationManagement) when a user grants consent to an application. To see these events, go to Entra ID > Enterprise apps, then Audit logs under Activity, with at least the Reports Reader role, and filter on that activity. For a standing list, sign in as at least a Cloud Application Administrator, open Entra ID > Enterprise apps > All applications, select an app and choose Permissions: Microsoft says the Admin consent tab shows permissions granted for the whole organization and the User consent tab shows permissions granted to a specific user or group. Microsoft notes you cannot revoke user consent from the portal; that needs Microsoft Graph or PowerShell. AI tools that connect to company mail, files or calendar, such as some note-takers and writing assistants, can surface here because they ask for permission to read company data.

Microsoft Defender for Cloud Apps: the Generative AI category

If you have a Microsoft Defender for Cloud Apps license, it adds a traffic-based view of the AI sites used on onboarded devices or seen in uploaded logs. Microsoft's steps are: in the Microsoft Defender portal, go to Cloud apps > Cloud app catalog and set the Category filter to Generative AI to see the AI apps in the catalog with their risk scores. Microsoft's guide also shows filtering the Discovered apps tab under Cloud apps > Cloud discovery by Generative AI, which lists the AI apps actually seen in your traffic. Microsoft also describes a policy with the condition Category equals Generative AI that alerts when a new AI app is used, which turns a one-off search into an ongoing check.

Discovery needs traffic data. Microsoft says the out-of-the-box integration with Microsoft Defender for Endpoint uses the network logs from onboarded devices, without extra deployment or traffic routing, and that data can take up to two hours to appear after you turn it on (Microsoft Defender portal > Settings > Endpoints > General > Advanced features > Microsoft Defender for Cloud Apps). The main prerequisites Microsoft lists are a Defender for Cloud Apps license, devices onboarded to Defender for Endpoint (Windows 10 or 11, or macOS with network protection turned on), and either Defender for Endpoint Plan 2 or Microsoft Defender for Business, which Microsoft says is also part of Microsoft 365 Business Premium. Business Premium meets only that last prerequisite; the Defender for Cloud Apps license is a separate item in Microsoft's list. Without the integration, Microsoft says discovery relies on logs you upload or send through automatic log upload.

Two limits matter. Microsoft says that by default Defender for Cloud Apps cannot discover apps that are not in its catalog, which it describes as over 31,000 apps; a brand-new AI tool may be missing until someone suggests it to Microsoft or you add it as a custom app. And it sees only devices or networks whose traffic it receives, so a personal phone on mobile data is invisible to it.

Google Workspace: OAuth log events and accessed apps

In Google Workspace, two Admin console views show which third-party apps, including AI tools, are connected to company accounts. The first is the accessed apps list: go to Security > Access and data control > API controls, choose Manage Third-Party App Access and select View list under Accessed apps. Google says this needs the Service Settings administrator privilege, lists third-party apps that have accessed Google data, shows how many users use each app and which Google service APIs (OAuth scopes) it uses, and can be downloaded as a CSV. Google says the list is updated 48 hours after a token is granted or revoked, so very recent connections may not show yet.

The second is the OAuth log events report, for a dated trail of who authorized what. Google's path is Reporting > Audit and investigation > OAuth log events, which needs the Audit & Investigation administrator privilege. Google says the log records when a third-party app is authorized to access Google Account data and identifies the user for whom access was authorized or revoked. Google says the report shows the last 7 days by default, so widen the date filter for a fuller history. Some event types, such as API call events, are available only in certain Enterprise, Education and Cloud Identity Premium editions, so check what your edition shows.

Browser extensions on managed devices

AI browser extensions are easy to miss because many never ask for a company sign-in; they read the page directly. For managed Chrome, Google's Apps & extensions usage report (Devices > Chrome > Reports > Apps & extensions usage, Chrome administrator privilege) lists the apps and extensions installed on enrolled Chrome browsers and ChromeOS devices, with install counts, permissions and, for public Chrome Web Store extensions, a third-party risk score. It covers Chrome browsers and ChromeOS only, not Edge or Firefox. Chrome browser reporting must be turned on first, Google says data can take up to 24 hours to appear, and the report can be exported as a CSV.

On Windows devices managed with Microsoft Defender, the browser extensions inventory in Microsoft Defender Vulnerability Management lists extensions installed in Microsoft Edge, Chrome and Firefox, with the devices and users that installed each one and a permissions risk level. Microsoft says this needs Defender Vulnerability Management Standalone or, for Defender for Endpoint Plan 2 customers, the Defender Vulnerability Management add-on, and that it covers Windows devices only. If you have neither, an inventory tool you already run, or a short manual check on a sample of machines, is the fallback.

Expense reports, card statements and asking people

Paid AI subscriptions can show up in expense claims before they show up in any log. Search the last six to twelve months of expense claims and company card statements for AI vendors and for words such as AI, GPT, assistant, transcription and note taker. Each hit is a tool someone valued enough to pay for, which makes it a good candidate for approval with limits rather than a ban.

Then ask. A short, no-blame survey that asks which AI tools people use, for which tasks and with which kind of data can surface tools that no log shows, including personal-account use on personal phones. Make clear that the goal is to find a safe way to keep the useful ones.

Some AI use can slip past the sources above. AI features switched on inside software you already pay for, note-taker bots that join calls without asking for access to company data, coding assistants inside code editors, and installed desktop AI apps or AI browsers may not show up in sign-in, consent or extension reports. Check the admin settings of the software you already license for new AI features, look at meeting attendee lists for note-taker bots, and use any installed-software inventory you already run.

Discovery paths side by side

This table sums up each vendor-documented discovery path, what it finds, what it misses and what the vendor lists as required. Each row is sourced to the vendor's own documentation, checked on the date shown. The non-vendor path, expense records and a staff survey, follows the table. Menus, editions and licensing change often, so confirm against the linked page before you buy or plan around a feature.

Ways to discover AI tools in use: coverage, gaps and vendor requirements
Discovery path	What it finds	What it misses	Requirements the vendor lists	Checked on	Source
Microsoft Entra sign-in logs	AI tools that sign in with work Microsoft accounts, by user and app	Personal accounts; tools with their own username and password; history beyond the retention period	At least Reports Reader; 7 days retention on Entra ID Free, 30 days on P1 and P2	2026-10-21	[1] [2]
Entra audit log 'Consent to application' and Enterprise apps > Permissions	AI apps users or admins have allowed to read company data, with the permissions granted	Tools that never ask for consent; browser extensions that read pages directly	Reports Reader for audit logs; Cloud Application Administrator or Application Administrator to review permissions	2026-10-21	[3] [4]
Defender for Cloud Apps, Generative AI category	Browser and cloud AI apps seen in traffic from onboarded devices or uploaded logs, with risk scores; alerts on new AI apps	Apps not in the catalog; devices and networks whose traffic it does not receive	Defender for Cloud Apps license; for endpoint discovery, Defender for Endpoint Plan 2 or Defender for Business	2026-10-21	[5] [6] [7]
Google Workspace accessed apps list	Third-party apps that have accessed Google data, with user counts and scopes; CSV download	Tools that never connect to Google accounts; tokens granted in the last 48 hours	Service Settings administrator privilege	2026-10-21	[9]
Google Workspace OAuth log events	Dated record of app authorizations and revocations, by user	Tools that never connect to Google accounts; some event types depend on edition	Audit & Investigation administrator privilege; API call events only in certain editions	2026-10-21	[8]
Chrome Apps & extensions usage report	Extensions on enrolled Chrome browsers and ChromeOS devices, with permissions and install counts	Edge, Firefox and other non-Chrome browsers; unmanaged browsers and devices; data can take up to 24 hours to appear	Chrome administrator privilege; Chrome browser reporting turned on	2026-10-21	[10]
Defender Vulnerability Management browser extensions inventory	Extensions in Edge, Chrome and Firefox on Windows devices, by device and user, with permissions risk	macOS and other platforms; unmanaged devices	Defender Vulnerability Management Standalone, or the add-on for Defender for Endpoint Plan 2	2026-10-21	[11]

Checked against the vendor documents in the source list on October 21, 2026. No prices are shown because they change often and vary by agreement; check the vendor's current licensing page.

Expense reports, card statements and a staff survey: finds paid subscriptions and tools no log shows, including personal-device use; misses free tools nobody mentions; no vendor requirement. This path is our own advice, not vendor documentation, so it has no source or checked-on date.

What to do with what you find

Put every tool you find into one approved-tools register: tool name, who uses it, company or personal account, what data goes in, and a decision. Our free AI acceptable-use policy template includes an Excel approved-tools register you can use for this. The list below is a rule of thumb for the first decision, not a vendor or regulatory requirement; adjust it to your own data classification.

Company-licensed AI tool on business terms, used with work accounts: Approve. Record the owner and the data allowed; keep it on the register.
Useful tool used with personal accounts for work: Approve with limits if a business plan exists and passes vetting, then move users to company accounts; otherwise retire and point people to the approved tool.
App with consent to read company mail, files or calendar that nobody reviewed: Review now. Check the permissions granted; revoke them if the app fails vetting, and route future requests through admin consent.
Browser extension that reads every page: Review now. Keep only if vetted; otherwise remove it and add it to your managed-browser blocklist.
Paid AI subscription on an expense claim: Vet it. If it passes, move it to a company account and billing; if not, stop reimbursing it and offer the approved tool.
Tool you cannot identify, or one that is new to the catalog: Investigate. Ask the users what it does and what data goes in before deciding.

Make discovery a routine, not a one-off

New AI tools and new AI features inside existing software keep arriving, so a single sweep goes out of date quickly. Repeat the log, consent and extension checks on a regular cycle, such as quarterly alongside your access review, and turn on ongoing alerts where you have them, such as the Defender for Cloud Apps new Generative AI app policy. Review the register and the acceptable-use policy at the same time so they stay in step.

Common follow-up questions

Can I see which AI tools employees use on their personal phones?

Not through company logs, unless the phone is managed or the tool signs in with a work account. Sign-in, consent, discovery and extension reports all depend on accounts, devices or networks you manage. A short, no-blame staff survey is the practical way to learn about personal-device use.

Do I need Microsoft Defender for Cloud Apps to find shadow AI?

No. It adds a traffic-based view of AI sites used on onboarded devices or seen in uploaded logs, but Entra sign-in and consent logs, Google Workspace OAuth reports, browser extension reports and expense reports all work without it. They each see a narrower slice, so use several together.

How far back can I see AI sign-ins in Microsoft Entra?

Microsoft lists seven days of sign-in and audit logs on Microsoft Entra ID Free and 30 days on P1 and P2, checked October 21, 2026. Microsoft says upgrading does not restore data that has already expired, so route logs to storage if you need a longer history.

How do I see which third-party AI apps are connected to Google Workspace?

In the Google Admin console, go to Security > Access and data control > API controls, choose Manage Third-Party App Access and open the Accessed apps list. Google says it shows user counts and the scopes each app uses, can be downloaded as a CSV, and updates 48 hours after a token is granted or revoked.

What should I do first after finding unapproved AI tools?

Record each one in an approved-tools register, then review the ones with the most access first: apps with consent to company mail or files, and browser extensions that read every page. Approve the useful ones that pass vetting on company accounts and retire the rest with a pointer to the approved tool.

Need help with this?

LAN Service Group helps small and mid-size businesses run an AI discovery sweep across Microsoft 365 or Google Workspace, build the approved-tools register from what it finds, and keep the acceptable-use policy and controls in step with it.

Talk to LAN Service Group LAN Service Group AI services → (888) 281-7243