What is an AI system inventory?
An AI system inventory is a maintained list of the AI tools, models and AI-enabled features an organization uses, with the owner, purpose, data, account type and risk level of each. It is the starting record for AI governance: a business cannot govern tools it does not know it has.
Full guide: AI governance: the approval register, owners and review cycle Service: AI Governance Consulting
An AI system inventory, sometimes called an AI register or AI use case inventory, records where AI is in use across the business. For a small company that is rarely a list of models it built. It is the chatbot subscriptions staff signed up for, the AI assistant inside the office suite, the note-taker that joins video calls, the AI features switched on inside the CRM or accounting package, browser extensions, and any agents or automations that call a model.
Each entry answers the same questions. What is the tool and who supplies it? Who in the business owns it? What is it used for? Which data goes in, and does the vendor train on it? Is it a company account under single sign-on or a personal one? What can it reach through connectors, and can it take actions or only read? How risky is the use, and when was it last reviewed? The NIST AI RMF Playbook describes an AI system inventory as an organized database of artifacts relating to an AI system or model, which may include system documentation, incident response plans, data dictionaries and contact information for the people responsible.
The inventory is a named expectation in the main frameworks. The NIST AI Risk Management Framework includes it as Govern 1.6: mechanisms are in place to inventory AI systems and are resourced according to organizational risk priorities. Under OMB Memorandum M-25-21, each U.S. federal agency other than the Department of Defense and the Intelligence Community must inventory its AI use cases at least annually. A related document is the model card, a short summary of a single model's intended use, performance and limits, proposed by researchers in 2018. A business that buys AI tools will mostly read vendors' model cards and system documentation and link them from its inventory, not write its own.
Reviewed by Jamie Baum, Head Implementation Manager, AI Practice. Our editorial standards.
Why it matters for a small business
One of the first things a client's security questionnaire or an insurer asks about AI is which tools you use and what data they see. A small business often cannot answer, because adoption happened one free trial at a time. An inventory turns that into a one-page answer. It is also how risky tools get found: a meeting bot recording client calls, a personal account holding company files. For a small firm it can be a spreadsheet, provided someone owns it and reviews it on a schedule.
How NetSys builds the inventory
NetSys builds the inventory from evidence, not from memory. Sign-in and app-consent logs, managed browsers and a short staff survey show which AI tools and embedded AI features are in use and the accounts they run on. The result is a register of approved tools and uses, each with an owner, an account type, a risk level and a data rule, and a review calendar that starts with the first scheduled check.
Discovery has limits, and we say where they are: personal accounts and unmanaged devices can hide use, and some records show that a service was used without revealing what was submitted. The register is part of our AI governance consulting, and a point-in-time version comes out of an AI security assessment.
Sources
Checked on November 22, 2026.
- NIST AI RMF Playbook: Govern (Govern 1.6, inventory of AI systems), NIST Trustworthy and Responsible AI Resource Center
- OMB Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust, Office of Management and Budget (The White House)
- Model Cards for Model Reporting (Mitchell et al., 2018), arXiv (Cornell University)
AI System Inventory: FAQs
What should an AI inventory include?
For each tool or use: the name and vendor, the business owner, the purpose, the data that goes in, the account type, the connectors and permissions it has, whether it can take actions, the vendor's data and training terms, a risk level and the date of the last review. Include AI features inside other software, not only standalone AI products.
How do you find the AI tools employees already use?
Combine technical evidence with asking. Sign-in and app-consent logs in Microsoft 365 or Google Workspace show AI apps connected to company accounts, managed browsers and device management show extensions and sites, expense reports show subscriptions, and a short survey fills in the rest.
What is the difference between an AI inventory and a model card?
An AI inventory is an organization's list of all the AI it uses. A model card is a document about one model, usually written by its developer, that describes intended uses, performance and limitations. An inventory entry can link to the vendor's model card or system documentation as supporting evidence.
Does a small business need an AI inventory?
Yes, in a simple form. If staff use AI tools, a list of which tools, which accounts and what data is the minimum needed to write a policy, answer a client or insurer and notice a risky tool. A spreadsheet with a named owner and a review date is enough to start.
More terms
Start with an inventory built from evidence.
Tell us how many people you have and which platforms you run, such as Microsoft 365 or Google Workspace. We will explain how we find the AI tools and features in use, what the register records for each one and how it is kept current.
