Skip to main content

​
What It Is

The AI Inventory is an Admin Panel tab that provides a centralized view of every assistant and application in the organization. For each AI asset, it displays key information including its name, type, description, status, AI Act risk classification, usage, owners, comments, and change history.
Panta Dokumentationen (DE) 1
The classification is derived by AI from the asset’s stated use case and then reviewed by a person. The Inventory is therefore more than just a register - it provides a transparent view of the AI applications an organization uses, and the associated risk classifications. Importantly, this classification does not indicate whether an assistant or application is compliant or non-compliant. Every application in the Inventory is considered compliant. The risk classification reflects the nature of the use case, as defined under the AI Act. Some use cases are inherently higher risk than others - for example, AI used to make or support decisions that affect people is considered higher risk than AI used for administrative or data processing tasks.

​
Why It Matters

As teams build their own AI assistants, organizations need more than a simple list of applications. The Inventory acts as a management hub, giving leaders visibility into the AI their staff is using, how each application is being used, and the associated use case. This enables organizations to oversee AI adoption, demonstrate governance, and ensure appropriate safeguards are in place.

Complete overview

See every AI assistant and application used across your organization in one place. Track key metrics, filter by type, status, or AI Act risk level, and quickly find the assets that matter.

AI-assisted classification

If an asset has not yet been classified, AI evaluates its use case against EU AI Act criteria and proposes a risk level.

Human decision

AI recommends a classification, but people make the final decision. Organizations can review, adjust, and approve the recorded AI Act risk level, ensuring accountability remains with the organization.

Traceability

Every AI asset includes owners, comments, and a complete change history. Discussions, classification updates, and decisions are recorded to provide a transparent audit trail.

​
The Three Risk Levels

AI applications are classified into three risk categories according to their intended use case, with each category carrying different regulatory obligations.

High Risk

AI that makes or significantly influences decisions about people: applicant screening, performance evaluation, credit and financial decisions, medical diagnosis support, access to essential services, or legal decision support. Under the EU AI Act this covers the domains listed in Annex III as well as safety components of regulated products. Strict obligations apply, including risk management, human oversight, technical documentation, transparency, and post-market monitoring.

Limited Risk

AI that interacts with people or generates content without making consequential decisions: chatbots, content and marketing assistants, email drafting, summarization, and translation. Transparency obligations are the main duty here, meaning users are told AI is involved and generated content is disclosed where required.

Minimal Risk

AI for internal productivity, automation, and data processing: code helpers, document formatters, workflow automation, data extraction, internal reporting, scheduling, transcription. Most systems fall here and generally carry no obligations beyond applicable law and voluntary best practices.

​
What The Inventory Does Not Promise

The AI Inventory is not legal advice and not a certification. Automatic classification is a proposal based on the described use case, and the notice in the detail panel says so. Whether a system genuinely qualifies as high risk under the EU AI Act, and which obligations follow, is the organization’s call, with legal support where needed.

​
Owners and Comments

Every AI Inventory asset includes Owners, Comments, and Change History sections to support collaboration and governance throughout its lifecycle.
  • Owners identify who is responsible for an App or Assistant. The creator is automatically recorded, and additional owners can be assigned to share responsibility. Once ownership is granted, owners can also assign the apps and assistants to teams.
  • Comments provide a centralized discussion thread where teams can capture context, ask questions, document decisions, and collaborate on the asset’s use and classification.
  • Change History records AI Act risk classification updates and also general comments, creating a transparent audit trail of how the asset has evolved over time.

​
Tips and Best Practices

  • Add many details when setting up your assistants as the classification reads that description. For apps, the PANTA OS team will do the backend set up and ensure all of the classification details are included.
  • Review every High Risk proposal personally and record the decision in the comments.
  • Assign an accountable owner to every asset, otherwise no one is responsible when a question comes up.
Fastest way in: set the All AI Act Risk filter, look at the assets with no classification, and work through the list in one sitting. After that the ongoing effort is small.

​
Help center

In the Admin Panel under the AI Inventory tab, alongside Analytics, Users, Teams, Integrations, Community Feed, Organization Settings, and Token Limits. Access depends on your role; see Roles and permissions.
The AI evaluates the asset’s described use case against EU AI Act criteria and proposes one of the three levels. The proposal is marked as AI-generated and is meant to be reviewed before it is finalized.
Yes. The level is a dropdown in the detail panel with Not set, High Risk, Limited Risk, and Minimal Risk. Every change is recorded in the asset’s Change History.
Status describes whether the asset is enabled on your dashboard. In Use describes whether it is actually being used.
Comments keep the discussion attached to the asset. Change History is the audit log of risk changes and comments. Together they show that a classification was reviewed rather than simply accepted.
Export generates a file containing the current inventory list, suitable as an attachment for internal reviews or external audits.
Last modified on August 6, 2026