Hello All,
We are currently working on supporting AI data sources and their catalogs within ServiceNow CMDB and have been evaluating the appropriate CI classes and data model in alignment with CSDM 5 and the AI Control Tower/AI Inventory capabilities.
Use Case
The objective is to import AI data sources and their associated catalogs into CMDB. The discovered catalogs are classified and assigned sensitivity levels, and these classifications are stored in ServiceNow along with the corresponding CIs.
The AI data sources we are planning to support are:
- Azure OpenAI
- Atlas Vector Search
Proposed CI Class Design
Based on our review of CSDM 5 documentation, AI Control Tower, and learning resources available through ServiceNow University, we analyzed the available CMDB classes and are considering the following structure:
| Data Source Type | ServiceNow CI Parent Class | ServiceNow CI Child Class | Relationship Type |
| Azure OpenAI | Azure AI Agent | Logical Datacenter | Hosted on::Hosts |
| Atlas Vector Search | AI Function | Logical Datacenter | Hosted on::Hosts |
For Azure OpenAI, the Azure AI Agent class is available and directly extends the AI Function class.
However, we did not find a dedicated class for Atlas Vector Search, so we are currently considering using the AI Function class for representing this data source type.
We would appreciate confirmation on whether this approach aligns with recommended CMDB modeling practices or if creating a custom table extending AI Function class would be more appropriate for such AI data sources.
AI Catalog Components
AI catalogs may consist of components such as:
- Models
- Prompts
- Datasets
- Vectors
From our understanding of AI Control Tower, similar components appear to be represented using the following classes:
- AI Model Product Model
- AI Dataset Product Model
- AI Prompt Product Model
- AI Content Product Model
Based on this, we are planning to use these classes to store the imported catalog components.
We would appreciate guidance on whether this is the recommended approach for storing AI catalog components and how these catalogs should ideally be associated with AI Function/Azure AI Agent records.
Specifically, should these catalogs be linked using CI relationships, or should the association be maintained through the Model ID reference field (which references the Product Model class)?
AI Inventory Data Storage
During our analysis, we also explored the AI Inventory capability and wanted to confirm which CMDB tables are used by AI Inventory to store AI data sources, particularly for services such as Atlas Vector Search.
Identifying AI Data Source Types
If AI Function class is used to represent multiple AI data source types, we would need insights on the recommended way to identify or distinguish the specific AI data source type (for example, Azure OpenAI vs Atlas Vector Search) within the CMDB.
Sensitivity Classification Storage
The AI catalogs include sensitivity classification details, which need to be stored in ServiceNow. Currently, for other supported CI classes, we store such classification details in a key-value table configured as a related list. However, this related list cannot be configured for the above AI-related CMDB classes.
Thank you in advance for your guidance.
