Chapter 1 · Lesson 1.2
Inventory the organisation’s AI systems
You cannot analyse what you do not know about. First step: list every AI system used, bought or developed, including those teams adopted on their own.
| Inventory column | Question to ask |
|---|---|
| System and version | Which tool, which model, which version? Does the provider change the model without notice? |
| Purpose and users | What for, by whom, on which people? |
| Data processed | What data goes in, which of it is personal or sensitive, where does it go? |
| Role of the organisation | Provider (we develop it or market it under our name) or deployer (we use it under our responsibility)? |
| Decision | Does the output inform a human decision or replace it? |
| Owner | Who in the organisation is accountable for this system? |
Where to look
- purchases and subscriptions (writing, transcription, recruitment and customer-relations tools);
- AI features added to existing software (email, office suite, business software);
- individual use of consumer assistants, sometimes with company data: this is called “shadow AI”.
The pitfall
An inventory done once by the IT department misses half the uses. Ask the business units, and update it with every new tool.
Key takeaways
Start with a living inventory of AI systems, with purpose, data, role and owner, including AI features hidden in software and “shadow AI”.
Source: ISO/IEC 42001:2023 (inventory and responsibilities in an AI management system); Art. 4 of the Regulation (AI literacy).