AI inventory in a spreadsheet or in a tool?
Updated: 2026-09
A spreadsheet is enough to create an AI register. It reaches its limits on distributed capture across departments, on the traceability of a classification, and on audit-ready evidence. With few systems in one place it suffices.
A register of the AI systems in use can be kept in a spreadsheet, and for small companies with few systems that is a reasonable choice. We say so explicitly, because the first step matters more than the tool used for it.
The limits appear in three places: when several departments must contribute, when the reasoning behind a classification must stay traceable, and when evidence is required in audit-ready form.
| Criterion | Innopulse | Alternative |
|---|---|---|
| Getting started | Immediate, free | Guided, with set questions |
| Distributed capture | A file circulates, versions drift | Team workspaces, one source |
| Reasoning behind classification | Must be written yourself | Classification with legal citations |
| External evidence | Export the spreadsheet | Audit-ready report, verifiable URL |
| Currency of the legal position | Yours to keep up | Maintained |
| Cost | None | Free plan, then by tier |
When an alternative is the better choice
You have few AI systems, all in one area.
One person has the complete overview and maintains it reliably.
No external evidence is required for customers or supervisors.
FAQ
Is a spreadsheet enough for the AI Act?
As a register it can be. What it does not deliver is a traceably reasoned classification and audit-ready evidence — both become relevant as soon as somebody asks.
What matters more, the tool or the first step?
The first step, clearly. An incomplete register in a good tool is worse than a complete one in a spreadsheet.
Can we migrate from a spreadsheet?
Yes, and that is the usual route. The existing list is the basis; classification and reasoning are added.
Who should keep the register?
A coordinating person with input from the departments. That distribution is exactly where a single file fails in practice.
AI Risk Check
Classification with legal citations makes it traceable why a system was assessed as it was — in a review the reasoning counts for more than the result.
Team workspaces reflect the reality that AI systems sit across departments and nobody alone knows them all.
Verifiable public URLs answer the customer question about compliance status without anybody digging out a document.