A business application uses Gen AI: how are others modeling this?

[Keith Woods]

I want to start identifying Business Applications that consume Gen AI.  How are others doing this?

My user base wants yet-another attribute.  I am moving towards a related Technical Service/Offering.  Is anyone tackling this yet?

[Ashok Sasidhara]

In my opinion, the following high level steps aligned with CSDM 4.0 recommendations should be helpful:

1. Define relationships for business applications with their corresponding application services

2. Define relationships of these application services with technical service offerings

3. Create information objects (if already not present) and define relationships of those with business applications. This will be an important step as data privacy considerations will be even more important when Gen AI is being used.

Make sure to define all these relationships by considering the relationships prescribed in the following URL as per CSDM 4.0:

https://docs.servicenow.com/bundle/washingtondc-servicenow-platform/page/product/csdm-implementation...

These steps which i mentioned above would be a good starting point. But you should have discussions with configuration manager, application owners, service owners, enterprise architects etc. so that you can understand more about how various CSDM 4.0 elements are configured in your environment and then finalize a list of action items with inputs from all key stakeholders.

[s4scott · Accepted solution]

Similar approach in my organization. 
1. Create a business application for the AI vendor product
2. Create Application Service CIs for the implementations

3. Create a technical service offering (when there are multiple application service CI for the product)
4. Relate the TSO or App Service CI to the consuming Application Service
5. I agree with creating and managing the Information Object to represent data types involved, especially if you are in a industry with regulatory oversight on data disclosure. 

I too would like to see more guidance from ServiceNow on this. AI is another potential path for company confidential data to make it's way to the open internet.

[mikkojuola]

Hello,

I recently found out that ServiceNow has released new AI-related CI classes:

• cmdb_ci_function_ai
  • AI SaaS applications deployed on public cloud platforms that offer scalable, on-demand services for machine learning, data processing, and AI-driven tasks. This provides flexible solutions without the need for on-premises infrastructure management.
• cmdb_ci_appl_ai_application
  • AI software applications that can run on various platforms such as Linux, Windows, Docker containers, or Kubernetes (K8) clusters. These platforms support diverse AI workloads, including machine learning models, data analytics, and intelligent services or AI-enabled applications

I could not find any information on how those should be used, or how they relate to CSDM, but I would imagine that you can start creating these records to identify the GenAI solutions that you're using and then creating CI relationships to Business Applications, or maybe Application Services, to identify which systems are consuming these AI functions or apps.

Here's a list (screenshot) of all "AI-related CI and Asset Classes" in Yokohama release after updating the CMDB CI Class Models plugin to 1.71.0 version.

If anyone has more information about these CI Classes or the related Asset Classes, I'd be very interested to learn more about them. Thank you.

Cheers,

--Mikko

[s4scott]

I too was excited to see Mark Bodman's post on LinkedIn about the class updates, and the subsequent release of the CMDB Class Models plugin. I was only a little surprised there is no meaningful documentation yet, so I reached out to my company's sales team. 

Specifically, my organization would like to implement AI model governance. So I am less interested in the components of the AI implementation, and more focused on the "service" level view. The desire is to collect an inventory of standalone and integrated AI models used, what application services they support/interact with, what data types they contain, and, most importantly, what group in the organization owns the model.

More to come.