Agentic AI

[thannagi]

Is there a meta model in Mega how to inventory Agentic AI ?

How this is associated with deployment model, application association?

How do we define risk for agentic AI?

[AlanB]

That's a good point, but those characteristics are really "something delivered by the service", so I'd be inclined to model that as Technical Capabilities delivered by the Software Technology. An icon or new MetaPictures for the appropriate STs/ITS/MS should be an easy configuration though.

That said, if I was a HOPEX customer, I wouldn't be rushing to customise the tool to accommodate AI, I'd be trying to work out which standard boundaries I can work within. If AI proves to stick around (it's been years since anyone's asked me about modelling "The Blockchain", for example), then you can be sure that it will be incorporated into the tool. At that point, you really don't want to be undoing a lot of customisations.

Again, I hope that helps.

[thannagi]

Thanks for your response. Use of Microservice or IT service may help. But it doesn't allow us define characteristics of Agentic Ai, such as "Proactive, Adaptable, Collaborative etc.".  From diagram perspective, we need an icon to distinguish from Microservice vs IT Service. 

[AlanB · Accepted solution]

Hi,

I'm not sure there's a need for special metamodels or methods to model Agentic AI. It's a Software Technology which is instantiated by either a Micro Service or an IT Service*. That means it can be treated the same way as any other Software Technology for ownership, risk, deployment, etc., mapping, and use in processes.

That's the advice I give to my clients, and it seems to have worked for them, so I hope it helps,

Alan

*You can instantiate it as an Application too, of course, but in a corporate environment, I would expect to see an AI engine as "A part of a thing" rather than a standalone, which assumes you're using it "raw" and without any additional safeguards. I attended an AI and Cloud Festival in Copenhagen earlier this year, and one of the key messages from speakers was the need for constraints around, and management of, AI engine use.