Home / Glossary / AI inventory
Glossary
What is an AI inventory?
An AI inventory is the list of the AI a company uses, with the owner, the access and the status of each item. It covers AI agents, MCP servers, skills, models and AI apps. The list is a record and it stops nothing. Its value comes from the decisions and the blocking that follow it.
By the AIBound team · Updated 17 November 2026
Key takeaways
- An AI inventory lists the AI a company uses, who runs each item and what it can reach.
- A list alone stops nothing, so each entry needs a decision and a rule.
- An inventory built from a survey goes out of date as soon as employees install something new.
Why an AI inventory matters
Auditors, boards and customers now ask which AI a company uses. An AI inventory is the answer. It also gives the security team a place to record a decision for each AI: allowed, under review or denied.
The list is the starting point. A high-risk agent on the list keeps running until someone blocks it. A good inventory leads straight to action, and the action is what lowers the risk.
How an AI inventory is built
An AI inventory is built from what is found on devices and in company systems. It is kept current as AI appears or changes.
- 1 Find the AI in use, including AI that employees installed on their own.
- 2 Record each item with its type, the person who runs it and the device it runs on.
- 3 Add what it can reach, such as files, repositories or company systems.
- 4 Record the decision for each item and act on the high-risk ones.
AI inventory vs an AI risk assessment
An AI inventory says what is there. An AI risk assessment judges how much harm each item could cause. The assessment needs the inventory, because nobody can judge an AI they have not listed.
| Question | AI inventory | AI risk assessment |
|---|---|---|
| What does it answer? | Which AI the company uses | How risky each AI is |
| What does it hold? | Each AI, its owner and what it can reach | The risks, their likely harm and the planned response |
| What comes next? | A risk assessment of each item | A rule to allow, ask about or deny the AI |
Sources
- NIST AI RMF Core · airc.nist.gov
Sources checked 19 November 2026
Examples
AI inventory examples
FAQ
Frequently asked questions
What is an AI inventory?
An AI inventory is the list of the AI a company uses. It names each AI agent, MCP server, skill, model and app, the person who runs it and what it can reach. It is the record that an auditor or a board asks for, and the base for every decision that follows.
What should an AI inventory include?
Each entry should name the AI and its type, the person who runs it, the device or system it runs on, and the data and systems it can reach. It should also show the decision made about it, such as allowed or denied, and the date that decision was made.
Is an AI inventory enough to manage AI risk?
No. An inventory is a list, and a list stops nothing. A high-risk agent keeps running after it is written down. Risk falls when each entry gets a decision and a rule that is enforced. The inventory tells the security team where to act, and the action does the work.
How does AIBound keep an AI inventory current?
AISeeker reads each employee device and updates what it finds as AI appears or changes, in real time. Each AI shows the device, the person and what it can reach. AIAssure then records every finding and decision as evidence for your GRC program, so the list reflects what is running now.