Skip to main content

Shadow AI agents are not shadow AI tools: how to build an AI agent inventory.

A person using an unapproved assistant is one problem. An agent acting alone, with nobody's name on it, is another. The inventory is the first control.

Product
Novie Control
Format
Guide
Reading time
5 min
Published
2 October 2026
Updated
21 October 2026
In short

Shadow AI agents are AI systems that take actions inside your business without anyone having signed them off. They differ from shadow AI tools, where a person uses an unapproved assistant, because the agent acts on its own and keeps acting after the person has gone home. An AI agent inventory is the register of every agent running, who owns it, what it can touch and what it costs. It is the first control that makes agents governable.

Shadow IT, shadow AI, shadow AI agents.

Shadow IT is software used without IT approval: a file-sharing account, a messaging app, a personal cloud drive. Shadow AI is the use of AI tools without approval or oversight: a personal chatbot account, a browser extension, an AI feature switched on inside a licensed product. Shadow AI agents are the third step. An agent has a goal, tools it can call, and permission to act, and it was approved by the person who built it, on the day they built it.

  • What it is. Shadow IT: unapproved software. Shadow AI tools: unapproved AI assistants used by people. Shadow AI agents: unapproved AI systems that act on their own.
  • Who acts. Shadow IT and shadow AI tools: a person, each time. Shadow AI agents: the agent, repeatedly.
  • Main exposure. Shadow IT: data stored outside control. Shadow AI tools: data processed outside control. Shadow AI agents: actions taken outside control.
  • Approval. Shadow IT and shadow AI tools: usually none. Shadow AI agents: the builder, once, with no scope.
  • What decays. Shadow IT and shadow AI tools: little. Shadow AI agents: the approval, as models and tools change.

The difference that matters is the last one. A tool does what a person asks. An agent does what it was set up to do, and when the model behind it is upgraded or its tool surface grows, it can do more than anyone approved without anyone changing it.

TakeawayA tool does what a person asks; an agent does what it was set up to do, and its approval decays as its model and tools change.

Why this is now a board question.

In 2026, IBM's Cost of a Data Breach Report, drawn from breaches at 602 organisations, found that security incidents involving shadow AI rose to 43 percent of organisations, from 20 percent the year before. Only 40 percent of organisations reported using access controls on AI models and data.

IBM, Cost of a Data Breach Report, 2026.

Agents are arriving ahead of the controls. In its Q2 2026 Pulse Survey, KPMG found 53 percent of large US enterprises had AI agents deployed and 18 percent were orchestrating multiple agents across workflows, double the previous quarter. Deloitte's 2026 State of AI in the Enterprise found 74 percent expect to be using agents by 2027 and 21 percent have a mature governance model for them.

KPMG, AI Pulse Survey, Q2 2026; Deloitte, State of AI in the Enterprise 2026.

Novamind Labs' view: shadow AI is an inventory problem before it is a policy problem. A policy tells people what they should not do. It has no way of noticing when an agent does it anyway.

TakeawayShadow AI incidents reached 43 percent of breached organisations in 2026, agents are arriving ahead of the controls, and a policy cannot notice an agent that ignores it.

How to build an AI agent inventory.

An inventory is a register, kept current, with one row per agent. Eight fields do the work.

  1. Name and owner. The agent, and the person accountable for it. A team name is not an owner.
  2. Purpose. What business outcome it exists to produce, in one sentence.
  3. Model and tools. Which model it runs on, and which tools or APIs it can call.
  4. Systems touched. Every system it can read from or write to.
  5. Data reached. The classes of data it can see, including personal information.
  6. Approval. Who approved it, when, and the scope they approved. If the scope is blank, the approval is a signature on a demo.
  7. Cost. What it costs per run and per month, attributed to a team and project.
  8. Stop. How it is paused, and who can do it.

TakeawayOne row per agent, with a named owner, its model and tools, what it touches, who approved what scope, what it costs and how to stop it.

How to find shadow AI in your company.

In a survey of 302 cybersecurity leaders fielded in March to May 2025, Gartner found 69 percent of organisations suspected or had evidence that employees were using prohibited public generative AI. Gartner predicts that by 2030 more than 40 percent of enterprises will have security or compliance incidents linked to unauthorised shadow AI, and advises regular audits for shadow AI activity.

Gartner, survey of 302 cybersecurity leaders, March to May 2025.

Nobody keeps a list of the AI nobody approved, so the inventory is assembled from the places it leaves marks.

  • Procurement and finance. Vendor contracts with AI features, card statements with model-provider charges, cloud bills with inference lines.
  • Identity. Service accounts and API keys created in the last twelve months, especially ones with no human named on them.
  • Vendors you already pay. Ask each which AI features they have switched on inside the product you license, and which of those can take actions.
  • The teams. Ask engineering, operations, marketing and finance separately what runs on their behalf. The gaps between the answers are the finding.
  • Sign-in grants. Your identity provider lists the third-party apps people have connected with their work accounts, including AI assistants and agent tools.
  • Network traffic. Web gateway and DNS logs show which AI services are reached from company devices, and how often.
  • Devices. Endpoint management shows installed browser extensions and desktop apps with AI features.

Run the search again every quarter. Each pass adds rows to the inventory and shows whether the governed path is getting easier to use than the workaround.

TakeawayNobody keeps a list of the AI nobody approved, so assemble it every quarter from invoices, service accounts, vendors, sign-in grants, network logs and each team's own answer.

Keeping it current.

An inventory that is right on the day it is built decays at the rate agents change. Two habits keep it alive. First, every new agent enters through the same governed path, so its row is written when it is built rather than discovered later. Second, when an agent's model, tools or reach change, its approval comes up for renewal and its row is reviewed. The approval is a statement about a capability at a point in time; store it that way.

TakeawayEvery new agent enters through one governed path, and every change to its model, tools or reach brings its approval up for renewal.

How Novie Control does it.

Novie ControlAvailable now

Governance and accountability.

Give every workload a named owner, a budget and clear boundaries.

1,284

calls governed today.

Sources

  1. Gartner, Gartner Identifies Critical GenAI Blind Spots That CIOs Must Urgently Address (19 November 2025)
  2. IBM, Every AI agent followed the rules, and the data still leaked (15 September 2026), on the Cost of a Data Breach Report 2026
  3. IBM X-Force, AI-powered adversaries and the enterprise risk challenge (29 July 2026), on the Cost of a Data Breach Report 2026
  4. IBM, IBM Study: One in Four Malicious Breaches are AI-Enabled, Costing Companies $6 Million on Average (29 July 2026)
  5. KPMG, AI Quarterly Pulse Survey Q2 2026 (24 June 2026)
  6. Deloitte, Business and IT leaders report AI agents are scaling faster than their guardrails (24 April 2026)

Other articles

Frequently asked questions

How do you find shadow AI in your company?

Look where it leaves marks: card statements and vendor contracts, API keys and service accounts with no named owner, apps connected through your identity provider, web gateway and DNS logs, browser extensions on managed devices, and AI features switched on in software you license. Ask each team separately what runs on its behalf, and record every find in an inventory with an owner.

What is the difference between shadow AI and shadow IT?

Shadow IT is software used without IT approval; the exposure is data stored outside control. Shadow AI is AI used without approval; the exposure is data processed, and increasingly actions taken, outside control. In 2026, IBM reported that security incidents involving unapproved AI rose to 43 percent of organisations, from 20 percent a year earlier.

How do enterprises govern AI agents in production?

By putting every agent on one governed path: a named owner, a scoped approval that renews when the agent changes, cost recorded per run and attributed to a team, an audit trail that answers what ran and who approved it, and a way to stop it. In 2026, Deloitte found 21 percent of organisations had a mature governance model for agents while 74 percent expect to run them by 2027.

How do I find the AI agents running in my organisation?

Start where they leave marks: vendor contracts and card statements, service accounts and API keys with no human named on them, AI features switched on inside licensed products, and separate conversations with each team about what runs on their behalf. Record each agent in an inventory with an owner, scope, systems touched and cost.

What should an AI agent inventory contain?

One row per agent with eight fields: name and owner, purpose, model and tools, systems touched, data reached, approval with scope and date, cost per run and per month, and how it is stopped. Review the row whenever the agent's model or tools change.

Build the intelligenceyour organisation compounds.