Problem
As organizations adopt AI agents, models, and generative AI services, most CMDBs have no record of what AI is actually running, who's accountable for it, or where it came from.
ServiceNow's AI Control Tower is built to close that gap, but its lifecycle, governance, and monitoring workflows depend on the same foundational CSDM entities — Product Model, Digital Asset, and Configuration Item — being populated with real ownership and provenance data. Without it, there's no responsible owner to notify when a model drifts or fails a fairness check, no vendor or manufacturer trail to support due diligence and EU AI Act or ISO/IEC 42001 reporting, and no link between an AI asset and the product model it's an instance of.
This Minimum Viable Data Model (MVD) blueprint establishes that baseline — the smallest set of entities and relationships needed to make an AI Digital Asset governable — before extending into AI Asset Type-specific attributes or deeper CI-level detail.
What the Blueprint Includes
The blueprint centers on AI Digital Asset and connects it to the product model it's based on, the companies that manufacture and supply it, and the people accountable for it. The Model relationship resolves automatically to one of two product model classes depending on the digital asset's actual type.
- AI Digital Asset – Root entity. Any AI-related digital asset — a model, dataset, prompt, agent, AI system, or MCP server — with an identifiable lifecycle and traceable ownership.
- AI Content Product Model – The product model for Model, Dataset, and Prompt-type digital assets (everything except AI System and MCP-type assets).
- AI System Component Product Model – The product model for AI System and MCP-type digital assets specifically; ServiceNow resolves the Model relationship here instead of AI Content Product Model when the asset's class is AI System Digital Asset or MCP Digital Asset.
- Provider – The manufacturer of whichever product model applies — Content or System Component — i.e. who built the underlying model or system component.
- Company – The vendor the AI Digital Asset itself is procured or licensed from.
- User – Assigned as Owner ("Owned by") and day-to-day manager ("Managed by") of the AI Digital Asset, both scoped to active users.
The Asset type (or Model category) is included in the Field Setup, and every AI Asset must have an asset type defined.
Considerations
This is intentionally a minimum viable model — it establishes provenance and accountability, not full CI linkage, even though it already resolves Product Model by asset type.
- Model routes to one of two Product Model classes automatically. The Model relationship is conditioned on the digital asset's actual class: AI System Digital Asset and MCP Digital Asset records resolve to AI System Component Product Model, everything else — including Model, Dataset, and Prompt-type assets — resolves to AI Content Product Model. This isn't a manual either/or choice; ServiceNow picks the path based on the record's class, so any new AI Digital Asset subclass you introduce needs to be aligned to whichever Product Model family it actually belongs to.
- Manufacturer follows both Product Model paths, not just one. Both AI Content Product Model and AI System Component Product Model carry their own Manufacturer link to Provider, while Vendor stays on the AI Digital Asset itself, pointing to Company. Keeping Manufacturer at the Product Model level rather than a single shared link means provenance stays correct no matter which path a given asset's Model relationship resolves to.
- Watch for the deprecated AI System Product Model table. An earlier ServiceNow table, AI System Product Model (
cmdb_ai_system_product_model), is deprecated in current instances — AI System Component Product Model is its replacement, and it's what this blueprint uses. If you're extending this model further, don't reintroduce the deprecated table. - Owner and Manager are both scoped to active users. "Owned by" and "Managed by" are separate fields on the AI Digital Asset, each filtered to active users only — populate both distinctly so accountability doesn't default to a single person or go stale.
- Field Setup is available on request. The blueprint's Field Setup calls out keeping Asset Type (
model_category) and Life Cycle Stage Status populated on the AI Digital Asset — ask if you'd like the full Field Setup as a table. - Anchor lifecycle stage to AICT's model. ServiceNow's AICT life cycle for AI product models and digital assets runs Ideation → Design → Operational → End of Life (with stages like Pilot, Chartered, Build, Available, and Pending Retirement); keeping Life Cycle Stage Status current on the AI Digital Asset is what lets AI Control Tower's lifecycle and monitoring workflows actually track it.
Benefits
Populated well, this blueprint turns AI assets from untracked shadow tooling into governed, CMDB-native records ready for AI Control Tower.
- Clear accountability. A populated Owner and Manager on every AI Digital Asset gives governance and risk workflows someone to notify when a model needs review, retraining, or retirement.
- Traceable provenance, resolved by asset type. The Model relationship automatically routes to AI Content Product Model or AI System Component Product Model based on the digital asset's actual class, and each carries its own Manufacturer link — so provenance data stays correct regardless of what kind of AI asset you're tracking, without anyone having to choose a path manually.
- Native fit with AI Control Tower. Because the blueprint is grounded in the same Product Model and Digital Asset entities AICT uses, populated data slots directly into ServiceNow's AI governance workflows without custom tables or integrations.
- A fast, extensible starting point. It's quick to implement and designed to grow — add deeper AI Asset Type-specific fields or CI links once the ownership and provenance foundation is solid.
- Reduced compliance and operational risk. Untracked AI assets are exactly the gap that leads to the compliance blind spots and inefficiency AI Control Tower is meant to close — this blueprint removes that gap at the data layer.
Related Information
This template is based on the AICT with CSDM v1 whitepaper, available in the ServiceNow Community.
Check other templates related to the AI Control Tower.
How to Get This Blueprint?
If you’re already a Data Content Manager Customer, you can download the Blueprint from the Knowledge Base. You will need your login credentials to access the blueprint download page.
