Lyzr Research · The AI Estate Inventory Playbook · November 2026

The AI Estate
Inventory Playbook.

How to find every agent your enterprise runs, reconcile what people say against what systems see, and keep the register true. Twelve hiding places, eight lenses and a ten day sprint.

Scroll to read

To CISOs, CIOs, heads of AI, internal audit and the people given the job of counting

A list of agents is not an inventory. A ledger is.

SAIDA PERSON TOLD YOUSEENA SYSTEM SAW ITSHOWNAN OWNER PROVED ITSHUTA CONTROL ENFORCES ITEACH STEP IS HARDER TO FAKE
01

Most inventories are lists of claims.

A spreadsheet that someone filled in from memory says what people believe. It carries no evidence, no date and no owner who has signed. It decays the week after it is made.

02

A ledger holds each claim against evidence.

Every line says how we know it exists, who answers for it, how much it can do and when we last checked. Auditors have worked this way for a century, and the same tests apply to agents.

03

You can build the first one in ten working days.

This playbook gives the places to look, the lenses to look through, the fields to record, the way to find owners and a day by day sprint. It ends with how to keep the register true.

This is the method

Every figure is cited. Most telemetry comes from security vendors and is labeled as such. Part 10 is the one place we write about Lyzr, and it is marked.

Lyzr Research, November 2026

The short version.

  • The ledgerTreat the register as a ledger of claims. Grade each line as said, seen, shown or shut.
  • The placesAgents hide in twelve places. Most teams watch about four.
  • The lensesNo single scan finds them all. Use eight lenses and merge them into one record.
  • The overlapReconcile said against seen. The gaps are Known, Stray, Ghost and Orphan.
  • The registerStart with six fields. Six complete beats eighteen half done.
  • The ownersWalk a ladder to find owners, offer amnesty, then tier by blast radius.
  • The sprintTen working days from scope to readout, with a date to count again.
  • The pulseAdd triggers, a front door gate and five vital signs, or the register decays.

What this covers

Ten parts, from the ledger to the first census.

01

What a register is

The ledger

Why a list fails, and the evidence grades that make a ledger.

02

Where agents hide

Twelve places

A field guide to the places a census must visit.

03

How to look

Eight lenses

What each discovery method sees, and what it misses.

04

Said against seen

Reconcile

Four classes of finding and a way to size the unseen.

05

What to record

Register design

Six fields first, IDs, states and design failures.

06

Who answers

Owners and triage

The owner ladder, amnesty and blast radius tiers.

07

Do it

The ten day sprint

Day by day, with roles and a worked example.

08

Stay true

Keep it alive

Triggers, the front door gate and vital signs.

09

What is asked

Frameworks and rules

NIST, ISO 42001, the EU AI Act and OWASP.

10

A vendor view

Where Lyzr fits

Labeled. Where a platform can help.

The argument in one picture

A register is a ledger of claims, not a list of names.

A list says an agent exists. A ledger says how we know, who answers for it, and how recently someone checked. The more evidence on a line, the harder it is to fake and the longer it stays true.

SAIDSEENSHOWNSHUTCLAIM IN THE REGISTERSupport triage agentContract review botSales enrichment flowDev MCP server on a laptopFinance close agentA REGISTER IS A LEDGER OF CLAIMS. EACH CLAIM NEEDS ITS EVIDENCE.

How to use this playbook.

If you areStart atTake away
A CISO or head of risk
Parts 01, 06 and 09
The grades, the tiers and the framework map
A platform or cloud lead
Parts 02, 03 and 05
Where to look, and what to record
Internal audit
Parts 01, 04 and 08
The assertions, the overlap and the pulse
A program lead running the count
Parts 06 and 07
The sprint and the owner ladder
An executive sponsor
Part 07 final slides
The one page readout
01

The ledger.

What a register is, and why a list is not enough.

This part sets out the one idea the rest depends on. A register is a set of claims, and a claim without evidence is only an opinion. It introduces four evidence grades and six audit tests.

THELEDGER.EVERY LINE CARRIES ITS EVIDENCE

Four grades of evidence

Said, seen, shown, shut.

Said means a person told you. Seen means a system observed it. Shown means a named owner confirmed it. Shut means a control enforces it. Report each count on its own, because a register that is all said is the weakest kind.

SAIDA PERSON TOLD YOUSEENA SYSTEM SAW IT
A scan, a log, an admin console
SHOWNAN OWNER PROVED IT
A named owner confirms and signs
SHUTA CONTROL ENFORCES IT
No credentials without a record
EACH STEP UP IS HARDER TO FAKE. COUNT THEM SEPARATELY.

What each grade looks like

Four grades, four kinds of proof.

01

Said

A survey answer, an email, a line in a spreadsheet. Cheap to get and quick to go stale.

02

Seen

A scan hit, a log line, an entry in an admin console. Objective, but silent about who owns it.

03

Shown

An owner has confirmed it, in writing, with a date. Slower, and it is what makes a record accountable.

04

Shut

A control makes the record a condition of working, such as no credentials without registration.

Our synthesis, offered as a starting point. The graphics are illustrative.

IBM Think 2026, via Alatirok

Fewer than one in five hold a complete inventory.

18%

Share of organizations reporting a complete AI inventory

Most organizations have something. Few have everything. The gap between a partial list and a full ledger is where unowned agents live.

Survey figure, as cited by a third party.

IBM Think 2026, as cited by Alatirok. See also the Cloud Security Alliance shadow AI note.

Six tests for a register

Auditors already know how to test a ledger.

Existence, completeness, rights, classification, valuation and cut-off are the assertions auditors apply to any record. They map one to one onto an agent register, so the same discipline works here.

01ExistenceIs it real andrunning today?02CompletenessIs everything onthe list, or onlywhat we know?03RightsDoes one namedperson answerfor it?04ClassIs it tiered bywhat it can do,not by name?05ValuationIs its blast radiussized and writtendown?06Cut-offIs the recorddated and stillcurrent?SIX TESTS AUDITORS ALREADY USE. APPLY THEM TO YOUR REGISTER.

The tests, in plain words · 1 of 2

The six tests, applied to agents.

01

Existence

Pick ten lines. Can you show each agent running today, with a log or a screenshot.

02

Completeness

Pick ten agents from a scan. How many are on the register. The miss rate is the answer.

03

Rights

Does a named person, not a team alias, answer for each line, and have they accepted it.

ISA 315 and ACCA audit assertions, adapted by Lyzr Research.

The tests, in plain words · 2 of 2

The six tests, applied to agents.

01

Classification

Are tiers set by what the agent can do, not by its name or its vendor.

02

Valuation

Is the blast radius written down for every agent above the lowest tier.

03

Cut-off

Is each record dated, and was it confirmed inside the review window.

ISA 315 and ACCA audit assertions, adapted by Lyzr Research.

The principle

“Count the evidence, not only the agents.”

THE LEDGER RULE

Report how many lines are said, seen, shown and shut. A register of 400 agents at 90 percent said is a smaller fact than a register of 120 agents at 80 percent shown.

Why lists fail

Why a list fails inside a quarter.

01

No evidence

Nothing on the line says how we know. The next person cannot tell a fact from a guess.

02

No owner who signed

A team name is not a person. When something goes wrong nobody picks up.

03

No date

Without a last confirmed date, a stale record and a live one look identical.

04

No trigger

Nothing forces an update when an agent is built, changed or retired.

05

No link to action

A register that cannot change an access decision is a document, not a control.

Our synthesis, informed by the Anomity and enz.ai inventory guides.

02

Twelve places.

Where agents hide in a real enterprise.

Agents are not only built in the AI platform. They are switched on in SaaS settings, wired up in automation tools, installed on laptops and granted access through personal accounts. This part is the field guide.

THEPLACES.TWELVE DOORS, ONE ESTATE

The twelve doors

Four doors are watched. Eight often are not.

Most teams monitor the cloud AI platform, the main SaaS suites and the low-code builder they bought. The other doors lead to agents that nobody reviewed, and they are where a census earns its keep.

01
Cloud AI platforms
02
Embedded SaaS agents
03
Low-code builders
04
Workflow automation
05
Coding assistants
06
Local MCP servers
07
Repos and notebooks
08
CI/CD pipelines
09
Browser and desktop agents
10
Personal OAuth grants
11
Vendor and contract AI
12
Data platform agents
FOUR DOORS ARE USUALLY WATCHED. EIGHT OFTEN ARE NOT.

Places one to six

Places one to six.

01

Cloud AI platforms

Agents built in a cloud console by one team, in an account central IT may not review.

BedrockVertexAzure AI Foundry
02

Embedded SaaS agents

Assistants and agents switched on inside tools the business already pays for.

CRMService deskOffice suites
03

Low-code builders

Business users assemble agents with drag and drop, often with broad connectors.

Copilot studiosAgent builders
04

Workflow automation

Automation flows with a model call in the middle and credentials in the settings.

iPaaSRPAZaps
05

Coding assistants

Assistants that read repos, run commands and call tools from a developer machine.

IDE agentsCLI agents
06

Local MCP servers

Tool servers running on laptops that give assistants access to files and services.

Config filesLocal ports

Our synthesis, offered as a starting point.

Places seven to twelve

Places seven to twelve.

07

Repos and notebooks

Scripts and notebooks that call a model API with a key in code or in the environment.

SDK importsNotebooks
08

CI/CD pipelines

Pipelines that call models to review, test, triage or deploy.

ActionsRunners
09

Browser and desktop agents

Extensions and desktop tools that act on a user’s behalf in other applications.

ExtensionsComputer use
10

Personal OAuth grants

A user links an outside agent to work mail, drive or calendar with one click.

Consent grantsToken scopes
11

Vendor and contract AI

AI that arrived in a vendor product update or a services contract, never reviewed as AI.

RenewalsSub-processors
12

Data platform agents

Agents inside warehouses and lakehouses that query and write data directly.

SQL agentsNotebook jobs

Our synthesis, offered as a starting point.

Clutch Security, December 2025, vendor data

In a modeled 10,000 person organization, about 1,500 people run MCP servers.

15.28%

Share of employees running MCP servers in Clutch’s model for a 10,000 person organization

Clutch analyzed more than 15,000 MCP deployments and modeled the result at about 3,056 deployments in one organization. 38% of those were unofficial, and 86% ran locally. Treat the exact figures as a vendor model, and the direction as the point.

Vendor data. Modeled, not measured in one customer.

Clutch Security, December 4, 2025. Vendor research, sample size beyond 15,000 deployments.

What the MCP data says

Most MCP servers are local, unofficial and on endpoints.

The shape matters more than the decimal. Local servers on laptops are invisible to cloud scans, which is why endpoint and code lenses are needed. Clutch also found 3% of published servers held valid hardcoded credentials.

Employees who run an MCP server (modeled, 10,000 staff)15.28%MCP deployments that are unofficial38%Deployments with a local architecture86%Deployments that sit on endpoints95%Published servers with valid hardcoded credentials3%CLUTCH SECURITY, DEC 2025. VENDOR DATA, 15,000+ DEPLOYMENTS.

Pluto Security, August 2026, vendor data

Across four customer environments the count of distinct MCP servers grew by 72% to 98% in three months.

288

Median distinct MCP servers per organization in Pluto’s sample

From May to August 2026, Pluto saw distinct MCP servers on endpoints grow 72% to 98%. The median organization had 288 distinct servers and the median endpoint had 3. Sample sizes are mostly not stated, so read this as direction.

Vendor telemetry, four environments.

Pluto Security, August 2026. Vendor telemetry.

Cyberhaven Labs, vendor data

Native AI apps on endpoints grew by about five times in a year.

+509%

Growth in AI native apps on endpoints, February 2025 to February 2026

Coding assistants grew 357% over the same year, across hundreds of thousands of employees. Endpoints are where a lot of the estate now lives, and they are the lens most programs lack.

Vendor data.

Cyberhaven Labs, 2026. Vendor telemetry.

Why each place is missed

Why each place gets missed.

01

Different owners

Cloud teams, SaaS admins, developers and procurement each own one door and none sees the others.

02

No shared vocabulary

Platform vendors call them assistants, copilots, flows, skills or servers. A scan for agents finds only one word.

03

Silent defaults

Some features arrive switched on in an update, so no project ever existed to be listed.

04

Personal accounts

Consent grants and personal plans sit outside the corporate tenant.

05

Nothing to approve

If setup takes a minute and needs no ticket, there is no ticket to find later.

Our synthesis.

03

Eight lenses.

How to look, and what each way of looking misses.

Discovery is a set of lenses, not a single scan. Each lens sees some doors well and others not at all. This part describes eight, with the signals to look for and the blind spot of each.

THELENSES.EIGHT WAYS OF LOOKING

Eight lenses, one record

Eight ways of looking feed one register.

Cloud accounts, SaaS consoles, code and CI, endpoints, identity, network egress, spend and people. Each produces a list. A census merges them into one record and notes which lens saw which agent.

ONE ESTATE REGISTERmerged and de-duplicated01Cloud accounts02SaaS consoles03Code and CI04Endpoints05Identity and OAuth06Network egress07Spend records08People and amnestyEIGHT LENSES, ONE MERGED RECORD.

Lens one and two

Lenses one and two, the platforms.

01

Cloud accounts

List agent, model and registry resources in every account and project. Include the accounts nobody on the central team created. Blind spot, everything off the cloud.

02

SaaS consoles

Export agents, assistants and automations from each admin console and builder. Ask each vendor what is on by default. Blind spot, tools without an admin API.

Our synthesis, informed by the Anomity discovery domains guide.

Lens three and four

Lenses three and four, the builders and the laptops.

01

Code and CI

Search repos for model SDK imports, API key patterns and workflow files that call models. Search for MCP config files such as .cursor/mcp.json, claude_desktop_config.json and .vscode/mcp.json. Blind spot, code outside your org.

02

Endpoints and browsers

Inventory installed AI apps, extensions and local servers through endpoint management. Look for listening local ports. Blind spot, unmanaged and personal devices.

Our synthesis. Config file names are common examples and vary by tool.

Lens five and six

Lenses five and six, the traffic and the grants.

01

Identity and OAuth

List app registrations, consent grants, service accounts and token scopes. Flag any grant to an AI vendor. Blind spot, agents that use shared secrets rather than identities.

02

Network egress

Match proxy and DNS logs to model provider domains and gateways. It reveals use even when nothing is installed. Blind spot, encrypted tunnels and personal networks.

Our synthesis.

Lens seven and eight

Lenses seven and eight, the money and the people.

01

Spend and procurement

Search card statements and purchase orders for AI vendors and usage billing. Spend finds what nothing else does. Blind spot, free tiers.

02

People and amnesty

Ask. A short survey with an amnesty promise finds agents that were never recorded anywhere. Blind spot, whatever people forget or hide.

Our synthesis.

What each lens sees and misses.

LensSees wellMissesEvidence grade
Cloud accounts
Platform built agents
Anything off cloud
Seen
SaaS consoles
Embedded and low-code
Tools without APIs
Seen
Code and CI
Scripts, keys, MCP configs
External code
Seen
Endpoints
Local apps, extensions, servers
Personal devices
Seen
Identity and OAuth
Grants, service accounts
Shared secrets
Seen
Network egress
Usage with nothing installed
Tunnels, personal networks
Seen
Spend
Paid tools of any kind
Free tiers
Seen
People and amnesty
Unrecorded builds
Forgotten or hidden
Said

How many lenses

Run at least four lenses, and always one of each kind.

  • One platform lens, from a cloud or SaaS console.
  • One builder lens, from code and CI.
  • One traffic or grant lens, from identity or egress.
  • One human lens, from people and amnesty.
PLATFORM LENSBUILDER LENSTRAFFIC OR GRANT LENSHUMAN LENS

Our recommendation. Independent lenses also allow the estimate in Part 04.

Scan hygiene

Five rules for scanning.

01

Read only first

Start with read-only access. A census that changes things loses trust on day two.

02

Write the definition first

Agree what counts as an agent before the first query, then freeze it for the sprint.

03

Log the query

Record each query and date so the evidence can be reproduced by an auditor.

04

Name the blind spot

State what each lens could not see in the readout.

05

Do not rate yet

Collect before you judge. Tiering is a later step.

Our synthesis.

The definition

One definition of an agent, written before you scan.

01

Autonomy

It can take actions or call tools without a human approving each step.

02

Context

It holds instructions, memory or retrieved data beyond a single prompt.

03

Reach

It is connected to at least one system, file or service that can change or leak something.

04

Edge cases

A chat window with no tools is out of scope. A workflow with a model call and write access is in.

Our synthesis. Adjust to your policy, and keep it stable during a count.

04

Reconcile.

What people said against what systems saw.

A census produces two lists. One is what people say they run. The other is what systems observed. The gaps between them are the findings, and they come in four kinds.

THEOVERLAP.WHAT SAID MEETS WHAT SEEN

Four classes of finding

Known, Stray, Ghost and Orphan.

Known is said and seen. Stray is seen but nobody said it. Ghost is said but never seen. Orphan is seen and recorded, with no live owner. Each class needs a different next step.

GHOSTSAID ONLYKNOWNBOTHSTRAYSEEN ONLYORPHANWHAT PEOPLE REPORTWHAT SYSTEMS FOUNDKNOWNMatch. Confirm owner and evidence.STRAYFind the owner. Register or retire.GHOSTVerify it exists. Close or locate.ORPHANSeen and said, no live owner. Assign.RECONCILE WHAT IS SAID AGAINST WHAT IS SEEN.

What to do with each class

A different move for each class.

01

Known

Confirm the owner, grade the evidence, tier it. Most of the register should end up here.

02

Stray

Find the owner using the ladder in Part 06. Then register it, or retire it. Strays are the main risk.

03

Ghost

Check whether it exists. It may be retired, renamed or running in a place no lens covers. Close or locate it.

04

Orphan

The owner left or changed role. Assign a new named owner within a set number of days or disable access.

Our synthesis.

Identity resolution

Joining the lists is the hard part.

01

Pick join keys

Use a stable ID where one exists, then the account, the workspace, the credential ID, the owner email and the endpoint name.

02

Expect many names

The same agent will be called different things by a cloud console, a repo and a person. Match on keys, not on names.

03

Keep the link

Store which lens and which record produced each match, so the result can be audited.

04

Mark uncertainty

Mark a match as probable or confirmed. Do not force a clean join.

05

Review the unmatched

The unmatched are where the findings are. Review them by hand.

Our synthesis.

Sizing the unseen

A way to estimate what you have not found.

If two lenses look independently, the overlap tells you how much each one misses. Method A found 40, method B found 30 and 12 appear in both, so about 100 exist and about 42 remain unseen. It is an estimate, and it is honest about the gap.

28ONLY IN A12BOTH18ONLY IN BMETHOD A FOUND 40METHOD B FOUND 30ESTIMATE OF THE WHOLE40 × 30 ÷ 12 = 100Seen by either method, 58. So about 42 remain unseen.Works only if the two methods look independently.ABOUT 42 STILL UNSEENA WAY TO SIZE THE UNSEEN. IT IS AN ESTIMATE, NOT A COUNT.

When the estimate holds

Use the estimate with care.

01

Independence

The lenses must look in different ways. Two scans of the same cloud are not independent.

02

Same definition

Both lenses must count the same thing, at the same time.

03

Heterogeneity

Easy to find agents are over-represented in both lists. That means the true total is likely higher.

04

Report a range

Say about 100, probably more. Do not print a single number as if it were a count.

Capture and recapture is a standard population estimating method. Adapted by Lyzr Research. An illustrative example.

Reco 2025, via the Cloud Security Alliance, vendor data

Shadow AI tools can run for more than a year before anyone detects them.

403

Median days to detect shadow AI tools in the cited telemetry

A single census pass is a snapshot. The longer an agent runs unseen, the more data and access it accumulates. This is why reconciliation has to repeat.

Vendor telemetry, as cited by a third party.

Reco 2025, via Cloud Security Alliance AI Safety Initiative, Shadow AI Apps, May 30, 2026.

A reconciliation worksheet, one row per finding.

FieldWhat goes in itExample
Finding ID
A new ID per row
F-0412
Class
Known, Stray, Ghost or Orphan
Stray
Seen by
Lenses that saw it
Code, egress
Said by
Who reported it, if anyone
Nobody
Next step
The class specific action
Find the owner
Due
A date, not a hope
Day 8

A good reconciliation

What a good reconciliation looks like.

  • Every finding has a class and a due date.
  • The count of unmatched on both sides is reported, not hidden.
  • Each join is explained and can be replayed.
  • The estimate of the unseen is stated with its assumptions.
SAIDA PERSON TOLD YOUSEENA SYSTEM SAW ITSHOWNAN OWNER PROVED ITSHUTA CONTROL ENFORCES ITEACH STEP IS HARDER TO FAKE

Our synthesis, offered as a starting point.

05

Register design.

What to record, and how to name it.

Registers fail from over-design more than from under-design. This part sets out the six fields that make a record useful on day one, the next fields to add, how to assign IDs and the states an agent passes through.

THEREGISTER.SIX FIELDS BEFORE EIGHTEEN

Six fields first

A register with six fields beats one half done in eighteen.

Name, owner, place, reach, autonomy and evidence. If you cannot fill these for an agent, you cannot make a decision about it. Add more fields only when somebody uses them.

01NameA unique ID anda plain name02OwnerA named person,never a team alias03PlaceWhere it runs,whose account04ReachData and toolsit can touch05AutonomyActs alone, ora human approves06EvidenceSaid, seen,shown or shutSIX FIELDS FIRST. ADD THE REST ONLY WHEN SOMEONE USES THEM.

The six fields in detail.

FieldWhat it holdsRule
Name and ID
A plain name and a unique ID that never changes
One ID per agent, issued centrally
Owner
A named person with a role
Reject entries with no owner
Place
Where it runs, which account or tenant
Include the credential it uses
Reach
Data stores, tools and external services it can touch
List by what it can do, not by what it was meant to do
Autonomy
Acts alone, human approves, or human in the loop for risky steps
One of three values, defined once
Evidence
Grade and date, with a link to proof
Said, seen, shown or shut

Fields to add second · 1 of 2

The next ten fields, in the order people ask for them.

01

Purpose and users

What it is for and who it serves, in a sentence.

02

Model and version

Which models it calls, and through whom.

03

Data classes

The sensitivity of what it handles.

04

Tools and MCP servers

Every tool it can call.

05

Human approval points

Steps that need a person.

Our synthesis.

Fields to add second · 2 of 2

The next ten fields, in the order people ask for them.

01

Vendor and contract

Who supplies it and under what terms.

02

Tier

Output of the triage in Part 06.

03

Last review

The date of the last confirmation.

04

Retire by

A date or a condition.

05

Links

Tickets, repos, runbooks, dashboards.

Our synthesis.

IDs and states

IDs never change. States do.

01

Issue one ID per agent

Do not reuse an ID, ever. A retired agent keeps its ID so history stays readable.

02

Keep external IDs as aliases

Cloud resource IDs and vendor IDs are aliases, since they change with migrations.

03

Use five states

Proposed, Active, Paused, Retired and Unknown. Unknown is for findings not yet classed.

04

Allow only listed moves

Proposed to Active needs an owner and a tier. Active to Retired needs a revocation record.

Our synthesis.

Test the register

Run the six tests on your own register before you publish it.

Existence, completeness, rights, classification, valuation and cut-off apply to the register itself. If it fails any of them, say so in the readout. A flawed register that admits it is more useful than a clean one that does not.

01ExistenceIs it real andrunning today?02CompletenessIs everything onthe list, or onlywhat we know?03RightsDoes one namedperson answerfor it?04ClassIs it tiered bywhat it can do,not by name?05ValuationIs its blast radiussized and writtendown?06Cut-offIs the recorddated and stillcurrent?SIX TESTS AUDITORS ALREADY USE. APPLY THEM TO YOUR REGISTER.

Design failures · 1 of 2

Six ways registers fail by design.

01

Defining AI too narrowly

Counting only models you built misses embedded and vendor agents.

02

One function owns it

A register owned only by security lacks the build knowledge. Joint ownership with engineering works better.

03

Waiting for perfection

Publish version one in the sprint. A rough register in use beats a perfect one in draft.

Failure modes drawn from the enz.ai inventory guide, a vendor source. The decay claim is that guide’s, not measured here.

Design failures · 2 of 2

Six ways registers fail by design.

01

Ignoring vendor AI

Contract AI is part of the estate even if nobody on your team built it.

02

Disconnected from action

If nothing in access, procurement or change management reads the register, it will decay.

03

No maintenance plan

One guide suggests an inventory may be materially incomplete within three to six months.

Failure modes drawn from the enz.ai inventory guide, a vendor source. The decay claim is that guide’s, not measured here.

06

Owners and triage.

Put a name beside every agent, then decide how much care each needs.

An unowned agent cannot be reviewed, retired or defended. This part gives a ladder for finding owners, an amnesty window, and a four axis blast radius score that sets the tier.

?THEOWNERS.A NAME BESIDE EVERY AGENT

Entro Labs, as cited by the Cloud Security Alliance, vendor data

Non-human identities may outnumber people by 144 to 1.

144 to 1

Ratio of non-human to human identities reported in one vendor dataset

Rubrik reports about 45 to 1 in another dataset. Whatever the exact ratio, most identities are not people, and ownership is not obvious.

Vendor data. Ratios vary widely.

Entro Labs and Rubrik, as cited by Cloud Security Alliance AI Safety Initiative, Shadow AI Apps, May 30, 2026.

The owner ladder

Walk down the ladder until someone answers.

Start with the creator of record, then the last editor, the cost center, the creator’s manager and the team lead. If none answer, the amnesty desk takes the agent and starts a clock. Never stop at a team alias.

RUNG 1Creator of recordSTRONGRUNG 2Last editorSTRONGRUNG 3Cost centerMEDIUMRUNG 4Creator’s managerMEDIUMRUNG 5Team leadWEAKRUNG 6Amnesty deskLASTWALK DOWN UNTIL SOMEONE ANSWERS. NEVER STOP AT A TEAM ALIAS.

Amnesty

Offer amnesty, and mean it.

01

The offer

For two weeks, anyone who registers an agent they built outside process will not be penalized for that.

02

The condition

After the window closes, an unregistered agent found by a scan is handled as a finding.

03

The message

Send it from a senior sponsor, not from security alone. State the benefit, which is support and safe credentials.

04

The channel

Use a one page form and a named inbox. If registering takes more than ten minutes, people will not do it.

Our synthesis.

Cloud Security Alliance and Oasis, vendor commissioned

About half of organizations say agent ownership is unclear.

51%

Share reporting no clear owner for AI agents or their identities

Ownership is not a data entry problem. It is a decision about who is accountable, and it needs a sponsor.

Survey, as cited by a third party.

CSA and Oasis, as cited by the Anomity inventory guide.

Blast radius

Score reach, power, autonomy and exposure.

Reach is how many systems and people it touches. Power is what it can change or spend. Autonomy is how often it acts alone. Exposure is what outsiders can reach or feed it. Score each from one to five.

ReachSYSTEMS AND PEOPLE IT TOUCHESPowerWHAT IT MAY CHANGE OR SPENDAutonomyHOW OFTEN IT ACTS ALONEExposureWHAT OUTSIDERS CAN REACHSCORE 1 TO 5, EACH AXIS. EXAMPLE SHOWN.PRIVATE DATAUNTRUSTED CONTENTEXTERNAL ACTIONLETHALTRIFECTA

The lethal trifecta

Three properties that should never meet unguarded.

01

Private data

The agent can read information that outsiders should not see.

02

Untrusted content

It reads text from email, web pages, tickets or documents that anyone can write.

03

External action

It can send, post, call or move something outside.

04

Why it matters

With all three, hidden text in untrusted content can make an agent send private data out. Remove one leg.

Simon Willison, The lethal trifecta for AI agents, June 16, 2025.

Tiers set the care, not the name.

TierTypical scoreControlsReview
Tier 1, low
Under 8
Registered, owner named
Yearly
Tier 2, moderate
8 to 12
Plus logging and scoped credentials
Every six months
Tier 3, high
13 to 16
Plus human approval on risky steps and kill switch
Quarterly
Tier 4, critical
17 to 20
Plus pre-launch review and continuous monitoring
Monthly

Using the tiers

How tiers change what happens.

01

Credentials

Higher tiers get shorter lived, narrower credentials.

02

Approvals

Tier 3 and 4 agents need a human to approve irreversible steps.

03

Monitoring

Tier 4 agents are watched continuously, and anomalies page a person.

04

Retirement

Any agent unused for 90 days moves to Paused, and its credentials are revoked.

Our synthesis.

Triage in one page

What the triage page must show.

  • Every agent, its tier and its owner.
  • The ten highest scores and who is acting on each.
  • Agents with no owner after amnesty, and the date they will be paused.
  • The count of strays and orphans open, and the age of the oldest.
AGENT, TIER, OWNERTOP TEN SCORESNO OWNER, PAUSE DATESTRAYS AND ORPHANS

Our synthesis, offered as a starting point.

07

The ten day sprint.

From scope to executive readout, in two working weeks.

A first census does not need a program. It needs ten days, six roles and a plan. This part sets it out day by day, with the output of each day and the trap to avoid.

THESPRINT.TEN WORKING DAYS TO A FIRST COUNT

Ten days at a glance

Ten working days. Every bar produces a file.

Scope on day one, pull lists from clouds and SaaS on days two and three, scan code and endpoints, reconcile on day five, run amnesty and owner notices across the second week and brief the executive sponsor on day ten.

D1D2D3D4D5D6D7D8D9D10Scope and definePull cloud and SaaS listsScan code and endpointsIdentity and egressMerge and reconcileOwner notices and amnestyTriage and tierExecutive readoutTEN WORKING DAYS. EVERY BAR PRODUCES A FILE.

Days one to three, scope and collect.

DayDoOutputTrap
1
Write the definition, name the sponsor, set the scope and roles
One page charter
Starting scans before agreeing the definition
2
Pull agent lists from cloud accounts and SaaS consoles
Two raw exports
Treating console counts as the whole
3
Scan repos for SDK imports and MCP configs, query endpoint management
Code and endpoint lists
Scanning only the main branch

Days four to six, widen and reconcile.

DayDoOutputTrap
4
Pull OAuth grants and service accounts, match egress logs to provider domains
Identity and egress lists
Ignoring grants to personal accounts
5
Merge, de-duplicate and classify as Known, Stray, Ghost or Orphan
Draft register and the finding list
Forcing a clean join
6
Send owner notices, open the amnesty window and the intake form
Notices sent, intake live
A message from security alone

Days seven to ten, close and report.

DayDoOutputTrap
7
Chase owners down the ladder, take the first confirmations
Owner status board
Stopping at a team alias
8
Close amnesty intake, score blast radius for all strays
Scores and tiers
Scoring before collecting
9
Finalize the register, run the six tests, estimate the unseen
Version one of the register
Hiding the gaps
10
Brief the sponsor with a one page readout and the next ninety days
Readout and plan
Presenting a number without its caveats

Who does what · 1 of 2

Six roles, none of them full time.

01

Sponsor

A senior executive who sends the amnesty message and clears blockers.

02

Program lead

Runs the sprint, owns the plan and the readout.

03

Cloud and SaaS admins

Pull the lists from the platforms they run.

Our synthesis.

Who does what · 2 of 2

Six roles, none of them full time.

01

Engineering lead

Scans repos and pipelines and answers for what the scans find.

02

Identity and network leads

Supply grants, service accounts and egress matches.

03

Risk or audit partner

Runs the six tests and signs off the method.

Our synthesis.

A worked example, composite

A composite sprint, in numbers.

Illustrative composite, not a real company. Four lenses produced 212 distinct candidates. 96 were Known, 71 Stray, 28 Ghost and 17 Orphan. After amnesty, 58 strays gained owners and 13 were retired. Evidence moved from 20% shown to 52% shown.

212 CANDIDATES, BY CLASS (COMPOSITE)96KNOWN71STRAY28GHOST17ORPHANSHOWN OR SHUT, BEFORE AND AFTER THE SPRINTBEFORE20%AFTER52%A COMPOSITE FOR ILLUSTRATION, NOT CUSTOMER DATA.

The composite, by class.

ClassCountWhat happened
Known
96
Owners confirmed, tiers set
Stray
71
58 registered after owner contact, 13 retired
Ghost
28
19 closed as retired, 9 located and registered
Orphan
17
11 reassigned, 6 paused pending an owner

The sprint traps

Five ways a sprint slips.

01

Debating the definition

Fix it on day one and revisit it only after the sprint.

02

Waiting for admin access

Request access before day one, not on day two.

03

Over-scoring

Scoring everything slows the week. Score strays first.

04

Skipping the readout

No readout means no decision, and no next count.

05

Treating it as one-off

Book the second census before the first one ends.

Our synthesis.

The executive page

The executive readout fits on one page.

01

The number

How many agents, in which classes, with the range for the unseen.

02

The grades

How much of the register is said, seen, shown and shut.

03

The risk

How many are Tier 3 or 4, and how many have no owner.

04

The ask

Two or three decisions needed from the sponsor.

05

The date

When the next count will be done.

Our synthesis.

08

Keep it alive.

A register is true on the day it is made. Then it starts to decay.

The first census is the easy part. This part covers what keeps the register true, which is update triggers, a front door gate, five vital signs and a plan for the registers that platforms now ship.

THEPULSE.A REGISTER THAT STAYS TRUE

Why inventories rot

Without triggers an inventory drifts away from the estate.

New agents appear weekly, owners change role and tools update themselves. A register held up by goodwill decays. One held up by triggers and a gate stays close to the truth. The curves are illustrative, not measured.

MONTH 0MONTH 3MONTH 6MONTH 9MONTH 12COMPLETENO UPKEEPTRIGGERS AND A FRONT DOOR GATEILLUSTRATIVE SHAPE, NOT MEASURED DATA.

Update triggers · 1 of 2

Six events that must update the register.

01

A new agent is built or bought

Registration is part of the build or purchase checklist.

02

An owner changes role or leaves

HR events trigger an ownership check.

03

A new tool or MCP server is added

A change in reach is a change in tier.

Our synthesis.

Update triggers · 2 of 2

Six events that must update the register.

01

A vendor ships an AI feature

Procurement flags renewals and release notes.

02

An incident or near miss

Re-score the tier and re-confirm the owner.

03

A scheduled recount

Run lenses again every quarter, and the full sprint yearly.

Our synthesis.

The front door gate

No registration, no credentials.

The most reliable control is one the builder needs. If an agent cannot get a credential, a key or a connector grant until it has a register entry, registration becomes the path of least resistance. This is how a record moves from shown to shut.

GATEIS IT REGISTERED?CREDENTIALSISSUEDNOT REGISTERED, NOT ISSUEDNO REGISTRATION, NO CREDENTIALS.

Building the gate

Where to place the gate.

01

Secrets and keys

Issue model and tool credentials only against a registered agent ID.

02

OAuth consent

Require admin approval for AI vendor consent grants, with a registration check.

03

Procurement

No purchase order for AI features without a register entry.

04

CI/CD

A pipeline that calls a model must reference a registered ID.

05

Exceptions

Allow a time limited exception with a named approver, and log it.

Our synthesis.

Five vital signs

Measure coverage, freshness, ownership, evidence and speed.

Coverage is the share of the estate with a record. Freshness is the share confirmed in the last 90 days. Ownership is the share with a live named owner. Evidence is the share at shown or shut. Speed is how soon a new agent is registered. Figures shown are illustrative.

72%Coverage
Share of estate with a record
85%Freshness
Records confirmed in 90 days
64%Ownership
Records with a live named owner
40%Evidence
Records at shown or shut
55%Speed
New agents registered in a week
FIVE VITAL SIGNS. VALUES SHOWN ARE ILLUSTRATIVE.

What each vital sign tells you.

Vital signDefinitionWarning sign
Coverage
Registered agents divided by estimated total
Falls as new places appear
Freshness
Records confirmed within 90 days
Drops after reorganizations
Ownership
Records whose owner is still in role
Drops when people move
Evidence mix
Share at shown or shut
Stuck at said
Time to register
Days from first use to a record
Rising means people route around the gate

Microsoft Agent 365

Microsoft documents a registry that sees more than Entra does.

01

Two views

The Entra admin center lists only agents that hold an Entra Agent ID. Agent 365 is described as the unified registry and control plane.

02

A broader inventory

Agent 365 inventory includes non-Microsoft agents and agents without an Entra Agent ID.

03

Where to look

Admins view it under Agents, All agents, in the Microsoft 365 admin center, with at least the AI Reader role.

04

The catch

Coverage depends on what is connected. Treat it as one feed, not the estate.

Microsoft Learn, Agent registry convergence, updated 2026. Vendor documentation.

The Register, April 9, 2026

AWS added a registry in preview across five regions.

5

Regions where AWS Agent Registry launched in preview through Bedrock AgentCore

It catalogs agents, tools, MCP servers and agent skills, supports MCP and A2A, and records can be added manually or fetched from configured endpoints. Preview features change, so check the current status before relying on it.

Vendor announcement, preview.

The Register, April 9, 2026.

A register of registers

Platform registries are feeds. Your register is the whole.

Treat each platform registry as one input, alongside code scans, identity and spend. Pull from them on a schedule, keep the source on each record and resolve duplicates against your own IDs. No single vendor sees the rest.

Microsoft Agent 365AWS Agent RegistryOther cloud consolesCode and MCP scansIdentity providerSpend recordsREGISTER OFREGISTERSPLATFORM REGISTERS FEED ONE RECORD. NONE IS THE WHOLE ESTATE.

Registry choices

Choosing where the system of record lives.

01

Use a platform registry

If most of your agents run on one platform, its registry is the natural first feed. Keep your own ID layer on top.

02

Use a governance platform

If you run many platforms, a neutral layer that ingests all the feeds saves manual joins.

03

Use a plain database

A governed spreadsheet or table is enough for the first six months if the fields are enforced.

04

Whatever you choose

Make it read by access, procurement and change management, or it will not stay current.

Our synthesis.

Operating model

Who runs the register after the sprint.

01

An owner for the register

One named lead with time and a mandate, not a shared inbox.

02

A monthly review

Thirty minutes on the vital signs, the open strays and the oldest orphan.

03

A quarterly recount

Repeat the four lenses and refresh the estimate of the unseen.

04

An annual sprint

Run the full ten days again, and compare with the last.

Our synthesis.

Failure modes · 1 of 2

Six ways a living register dies.

01

The champion leaves

Nobody else has the mandate.

02

The gate has a back door

People find a path around registration.

03

Triggers are manual

No system tells the register that something changed.

Our synthesis.

Failure modes · 2 of 2

Six ways a living register dies.

01

Metrics are vanity

Counting records instead of confirming them.

02

The register is private

If only security reads it, nobody else updates it.

03

Scope creep

Adding fields until updating takes an hour.

Our synthesis.

09

Frameworks and rules.

What the standards and the law ask you to know.

A register is not a compliance exercise, but it answers many compliance questions at once. This part maps the register to NIST, ISO 42001, the EU AI Act and the OWASP Top 10 for Agentic Applications.

THERULES.TEN RISKS, ONE REGISTER

NIST AI RMF, Govern 1.6.

PointIn practice
What it says
GOVERN 1.6 concerns mechanisms to inventory AI systems, resourced according to risk priorities. This is a paraphrase, and the exact text is on the NIST site.
What the register gives
A single list, with tiers that set effort by risk, which is what the control asks for.
What to show an assessor
The register, the method, the tiers and the date of the last recount.

ISO 42001 Annex B, resource documentation.

ControlTopicRegister field
B.4.2
Resource documentation
Name, purpose, links
B.4.3
Data resources
Reach, data classes
B.4.4
Tooling resources
Tools and MCP servers, model
B.4.5
System and computing resources
Place
B.4.6
Human resources
Owner, approval points

The EU AI Act dates a register helps with.

ObligationDateHow a register helps
General purpose AI rules
In force since August 2025
Shows which models you use and from whom
Article 50 transparency
August 2026, with a grace period to December 2, 2026
Finds user facing agents that need disclosure
Annex III high risk systems
Delayed to December 2027
Identifies candidate high risk uses
Annex I product systems
August 2028
Flags embedded AI in regulated products

OWASP in one view

Ten agentic risks, each one needs a register field to address it.

OWASP published its Top 10 for Agentic Applications in December 2025, with input from more than a hundred researchers. Most of the ten are easier to manage when the register says what an agent can reach and who owns it.

ASI01Goal hijackASI02Tool misuseASI03Identity and privilegeASI04Supply chainASI05Code executionASI06Memory poisoningASI07Inter-agent commsASI08Cascading failuresASI09Trust exploitationASI10Rogue agentsOWASP TOP 10 FOR AGENTIC APPLICATIONS, DECEMBER 2025.

OWASP risks one to five, and the field that helps.

RiskNameRegister field that helps
ASI01
Agent Goal Hijack
Reach, autonomy
ASI02
Tool Misuse and Exploitation
Tools and MCP servers
ASI03
Identity and Privilege Abuse
Owner, place, credential
ASI04
Agentic Supply Chain Vulnerabilities
Vendor, model and version
ASI05
Unexpected Code Execution
Place, autonomy

OWASP risks six to ten, and the field that helps.

RiskNameRegister field that helps
ASI06
Memory and Context Poisoning
Data classes, reach
ASI07
Insecure Inter-Agent Communication
Links, tools, owner of each side
ASI08
Cascading Failures
Tier, dependencies
ASI09
Human-Agent Trust Exploitation
Approval points, users
ASI10
Rogue Agents
Existence test, evidence grade, retire by

One register, many answers

What one register answers.

  • Which agents does a regulator or customer ask about, and who owns each.
  • Which controls apply to which tier, and the evidence they hold.
  • Which risks from the OWASP list each agent is exposed to.
  • Which agents moved this quarter, and why.
WHO OWNS EACHWHICH CONTROLS APPLYWHICH RISKS APPLYWHAT MOVED

Our synthesis, offered as a starting point.

10

Where Lyzr fits.

A vendor view, clearly labeled.

Everything before this part is vendor neutral and you can run it with a spreadsheet. This part is the one place we write about Lyzr, and what we think a platform can add.

WHERELYZR FITS.A VENDOR VIEW, LABELED

Where a platform helps and where it does not.

PointIn practice
Where it helps
Continuous discovery across clouds and SaaS, an agent register that is kept current, and the front door gate that links registration to credentials.
Where it does not
Defining the policy, finding owners, running amnesty and holding the sponsor conversation. Those stay with your people.
What to ask any vendor
Which of the twelve doors do you see, which grade of evidence do you produce, and how do you handle the agents you did not build.

Lyzr Opencontroller, one data point

In one enterprise, Opencontroller surfaced 227 shadow assets.

227

Shadow assets surfaced in one enterprise by Lyzr Opencontroller

One data point, not an average. Your number will differ, and a census will tell you what it is. We include it only to show how large the gap between known and actual can be.

Single customer observation. Not a benchmark.

Lyzr, internal observation.

Free Lyzr tools that pair with this playbook.

Tool or guideUse it forLink
Shadow AI Discovery Checklist
The first sweep on day one and two of the sprint
lyzr.ai/template/shadow-ai-discovery-checklist
AI Agent Sprawl Audit
A quick read on how large the gap might be
lyzr.ai/assessment/ai-agent-sprawl-audit
Agent Finder Tool
Finding agents the list does not show
lyzr.ai/assessment/agent-finder-tool
Agent Governance Maturity Assessment
Placing your register on the ladder
lyzr.ai/assessment/agent-governance-maturity-assessment
Reconcile your Kubernetes clusters against your agent inventory
The reconcile step in Part 04, for clusters
lyzr.ai/blog/reconcile-kubernetes-clusters
AI Agent Registry vs Model Inventory
Why a model list is not an agent register
lyzr.ai/blog/ai-agent-registry-vs-model-inventory

About Lyzr

The full-stack agent infrastructure platform for enterprises and governments that want to own their intelligence.

3,000+
Organizations building on Lyzr
40+
Enterprise customers
50+
Implementation partners
100,000+
Agent builders

Founded in New York in 2023. Agent Studio, Architect, Opencontroller, Open Cowork, Openfinetuner, Optimus and Madison cover building, governing, running and owning agents. Deploys in your AWS, Azure or GCP account, a dedicated tenant, SaaS, on bare metal, fully air-gapped, or on the Optimus appliance.

Backed by Rocketship.vc, BGV, Accenture Ventures, GFT, the Partnership Fund for New York City and Plug and Play.

11

Run the sprint.

Count first. Then name. Then repeat.

Pick a sponsor, book ten days and write the definition on day one. You will have a first count, a name beside every agent you found and a date for the next one. That is a governed estate, at version one.

COUNTFIRST.THEN NAME, THEN REPEAT

Your next step

Take the playbook. Run the sprint. Leave with your number.

Thirty minutes, no slides. Walk through your twelve doors, pick four lenses and set the date for day one.

01

Scope it

We agree the definition and the sponsor.

02

Look

We choose four lenses and plan the read-only pulls.

03

Name it

We plan the owner ladder and the amnesty window.