Shadow AI and agent discovery

Know which shadow AI tools your staff already use, who owns each, and what it can reach.

You get the list: the AI tools and agents found in your own records, who owns each, what data it can reach, and a costed plan to close the gaps. Read-only, and nothing is changed during the review.

What you get: An evidence-based AI tool and agent inventory, visibility gaps and a prioritized correction plan.
How it starts: A free scope call. The price is fixed in writing before anyone is given access.
When it is done: Discovered tools have an owner or an explicit ownership gap; the readout and costed options are delivered and review access is removed. Making the corrections is separate.

Price the shadow AI audit Review scope & terms

Illustrative sample · constructed example

Show the discovered tool and the gap beside it.

Read-only discovery · available evidence

Selected check

Is there enough evidence to assign a decision to the discovered tool?

AI use review / coverage recordIllustrative

AI use review / coverage record: constructed inspection record
Record	Evidence or responsibility	Sample review
Observed application	Record the discovered app and available evidence source.	Application observed
Permissions	Establish granted access and its business purpose.	Owner unknown
Unmanaged devices	State what the available evidence cannot establish.	Coverage gap remains
Action owner	Separate an accepted finding from a production control change.	Decision to assign

This is discovery and review. The record does not block a tool or establish complete AI-use coverage.

Evidence: Application grant recorded; accountable owner and purpose unresolved.
Decision: Investigate
Next action: Ask the client to identify the owner and validate the business use.

Illustrative sample. Available tenant and network records cannot establish an inventory of every AI tool. The review does not block access.

Open the complete illustrative document to inspect its fields and print a copy.

Open the sample document

Why now

The work starts here.

Shadow AI is in use across the business: tools staff signed up for with work accounts, agents running from laptops, and nobody holds the list.

Scope at a glance

One Microsoft 365 tenant
Read-only throughout
Tenant evidence, plus any discovery export you can provide

If nobody acts

What waiting costs

A consent granted in one click can hold standing access to mail or files, long after the person who granted it forgot the tool exists.
Without a list, there is no line between the tools you approved and the shadow AI you did not. Every policy conversation starts from a guess.
Data pasted into an unapproved tool is governed by that tool’s terms, not yours, and nobody can say which data has already crossed.
The first complete inventory usually gets assembled during an incident, under the worst possible conditions.

Check your own environment

Sounds like you?

Staff mention AI tools in passing that IT has never heard of
The enterprise-application list is full of names nobody recognizes
The board has asked what AI is in use and nobody could produce the list
The last attempt at an inventory was a survey, not tenant evidence

Save yourself the call

Not for you if

You already have an AI policy and a current, evidence-based list of what is approved and what is blocked
The only tool in question is Microsoft Copilot, which has its own engagement
The decision is already to block everything, which is a policy change rather than a discovery job

The work

What is different when we hand over

AI apps, agents and connected tools visible in the available evidence are inventoried: app consents and their scopes, enterprise apps, the Teams app list, software Intune has seen, and the audit log.
If the tenant already licenses Cloud App Discovery, the client exports a snapshot and it is folded in. If not, the report says what that would have added.
Each tool is recorded with its users, the data it can reach, a risk rating, and an owner, or a flag where no owner can be found.
Every entry gets a recommended allow, watch or block decision, with the reason, for the client to confirm.
The readout walks the list and the correction plan. Each fix is costed, so the follow-on can be scoped from it.

Acceptance

Checks before handover

Done when every tool found is on the list with an owner or a flag and a recommended decision, the readout is delivered, and access is removed.

Every in-scope discovered tool is in the inventory with its evidence source, users, reach and risk rating
Each entry carries an owner or an explicit flag, and a recommended disposition with the reason
The readout and the costed correction plan are delivered and access is removed

A fair question

Can we not just buy a shadow AI governance platform?

The platforms are good, and you may already license one inside Microsoft 365. Each has a blind spot: browser-based visibility needs an extension on every device, network discovery needs the logs, and an unmanaged laptop on home Wi-Fi is invisible to all of them. This audit tells you what your licenses already cover, what is switched on, and what none of them can see. If the honest answer is to switch on what you own, the report says so. AZ Innovations audits your tenant; it does not sell you a platform.

Price and calendar

Scope, price and project calendar

Engagement price

Price fixed after a short scope call.

The price is agreed in writing before any work starts and before anyone is given access. The call itself costs nothing. What moves the number:

What determines the scope

Size of the app-consent and enterprise-application surface
Evidence sources available: audit log retention, Intune enrolment, Teams app inventory, any Cloud App Discovery export
Number of users and teams in scope
Whether a board-level summary is required alongside the technical readout

Delivery calendar

The proposal names the start date, the change windows and the completion date before anything is agreed.

How engagements work

Scope & responsibilities

What is and is not included

The exclusions, what we need from you, how change is staged and reversed, and what is handed over at the end.

Exclusions

No production changes: revoking consents, blocking tools and configuring an approval path are quoted separately from the correction plan
Browser- and network-level discovery beyond tenant evidence, which needs client-side tooling or the Cloud App Discovery export
Employee monitoring or surveillance
Legal review of tool terms
Ongoing governance operation, which is handed to internal IT or the MSP

What we need from you

Grant read-only access to the tenant for the audit window
Export a Cloud App Discovery snapshot if the tenant licenses it
Confirm or change the recommended disposition on each discovered tool

Change and rollback method

The audit is read-only. Nothing is revoked, blocked or reconfigured during it.
No software is installed on user devices and no user is interrupted.
Access is granted for the audit and removed at the readout, with written confirmation.

What you keep at handover

A System guide: the configuration as delivered, the checks that were run and the unresolved list
The shadow AI inventory, with the evidence source recorded against every entry
The disposition register: recommended allow, watch or block, with owner or flag and reason
The consent and reach map: which tools hold which scopes into mail, files, chat and calendars
The costed correction plan handed over at the readout

Access, acceptance and support arrangements

How you know the work is good

Judge it by what you can inspect.

The method is written down, the sample records open and the demo can be broken on purpose. Read it the way your technical owner would.

How a project runs: A first conversation, free. An agreed scope with a named technical lead in the proposal. The price fixed in writing before we are given access. Checks before handover.
How we work
What is checked: Release checks, access test matrices and ownership maps. Constructed inspection records on the service pages show the standard before you buy.
See sample deliverables
What you keep: A System guide: what was agreed, what changed, what was tested and what remains open. With it, the scripts and check results behind the work.

Next step

Start with a free scope call.

Tell us what needs to change, what must keep working and the deadline. The call is free, and the price is fixed in writing before anyone is given access.

Price the shadow AI audit

Or write to [email protected]