AI and AICT in CSDM 5.0 Formal Guidance from ServiceNow

[scott_lemm]

This article is brought to you by the letters A and I ...

• How do we model AI in CSDM?
• What is an AI Asset?
• Where to I put my Agents?

Hot off the press... here is a summary overview of the ServiceNow AI Platform in the context of Enterprise and Business Service Management data modeling guidance provided by the Common Service Data Model (CSDM). 

This brief overview of both CSDM 5 and AI Control Tower (AICT) provides guidance on implementation to support better together life cycle service management within your ServiceNow Platform of solutions.

AND.... One more thing...

We provide 3 sample use cases to help spark understanding and discussion around AI in the CSDM.

NOTE: new entities are discussed:

• New Life Cycle State/Status (cancelled and deployable)
• New Digital System CMDB Class
• Re-introduce DevOps Package (Package) CMDB Class

Lets get AI Modeling!!!!

[SK Chand Basha]

Very Helpful!

[Bhausab Parama1]

Very Helpful!

[Dhruvii]

Thank you for writing this.

It is the clearest artifact I have seen on how AICT is meant to sit inside CSDM, and the grounding in Product Model, Digital Asset and CI across both the PoC and production lifecycles resolved several things I had been going back and forth on. The three reference blueprints in particular were worth the read on their own.

A few questions came up while I was mapping this against our own implementation, and I would appreciate any clarification:

1. AI Function scope. The Service Delivery section describes AI Function "and by extension MCP Function and AI Model Deployment" as the operational CI entities, while the summary data model describes AI Function as representing the operational CIs including AI Services, AI Agents and AI Model based services. Is AI Function intended as a parent class that MCP Function and AI Model Deployment extend, or as a peer class alongside them with the phrasing being shorthand for the layer as a whole? The class naming suggests the latter, and I want to make sure I model the hierarchy correctly.

2. Asset to CI traceability path. The summary data model links Digital Asset directly to AI Function with Deployed As and Deployment Of, but the blueprints route that traceability through the Package (Artifact). Are both paths expected to exist, or is one the authoritative link with the other being illustrative? Related to this, does the Package layer assume DevOps Change Velocity is in play, given sn_devops_package is the table shown?

3. Business Capability linkage shape. Blueprints 1 and 2 relate Business Capability directly to Business Application, while blueprint 3 and the summary model route through Digital System. Blueprint 3's note suggests Digital System is introduced when there are multiple Business Applications to group. Is the guidance to collapse Digital System when there is only one application, or to model it consistently regardless so that capability rollups have a single traversal path? I am inclined toward the latter, but would rather follow the intended pattern.

4. Non hyperscaler AI platforms. The paper notes that discovery can populate the AI CI entities. For organizations running their own agentic platform rather than a hyperscaler service, is the expectation that we feed those CIs through IRE as a custom discovery source, and are the identification rules for the shipped AI CI classes flexible enough to accommodate that?

Thanks again for putting this together. Happy to share what we find on our side as we work through the implementation. @scott_lemm

[Sean Cosgrove1]

@scott_lemm @markbodman Nice! As always, you guys rock at being on the forefront of services and data modeling. Looking forward to layering this into our current CSDM and AI initiatives!

[fkhondker]

Why does this contradict all the information present everywhere else? Specifically how AI application cmdb_ci_appl_ai class is defined vs the AI function class cmdb_ci_function_ai

[fkhondker]

Whelp, I read two white papers from servicenow and they contradict each other. 

Using CSDM 5 white paper w links white paper, our original definitions for the AI function table and the AI Model & application table are correct. AI function table is for SaaS applications on public cloud platforms, and AI Application table is for self hosted/Disney managed Applications aka Linux, Windows, Docker containers. We're having two buckets based on the hosting model. 

The one contradicting this definition is AICT with CSDM v1 from the article. This is the most recent one  and was released 2 weeks ago.  It's saying that AI Function and AI Application are two layers that stack on top of each other for the same AI system, regardless of hosting. AI Application is the installed software package(used if applicable), otherwise we always use the AI Function as the operational runtime underneath it. And specifically, it says AI Function "can represent both ServiceNow internal, Customer Private and Public Cloud service CIs" meaning hosting location doesn't determine which class you use.

So this will affect if we make 2 CI records for an AI system(to show two layers), or just create one based on how it's hosted. 

Sooo, which definition do we follow?