The Zurich release has arrived! Interested in new features and functionalities? Click here for more

Ask the Experts: Agentic AI Use Cases

Published on by Allen Andreas | Updated on

Join Ask the Experts host Allen Andreas as he welcomes MVP John Zhang who will conduct an interactive session highlighting real-world Agentic AI use cases.

ATE - 5_27_25 - Agentic AI Use Cases.png

Featured Guests

Kilo Patron

Event has ended

You can no longer attend this event.

Start:

Tue, May 27, 2025 08:00 AM PDT

End:

Tue, May 27, 2025 09:00 AM PDT

5,549 Views

6 Comments

What exactly is the difference between a chatbot or a well-built flow that is set to handle various tasks...and AI Agents? Aren't they basically the same thing? Unsurprisingly, this question comes up a lot!

With the ServiceNow Yokohama release, AI Agents were no longer something you just talked about...they arrived! During the Knowledge conference that just ended a few weeks ago, one particular experience that had a VERY long line all three days was 'Build an AI Agent', which introduced participants to AI Agents.

Of course, not everyone got to go to Knowledge 2025 and even if you did, you may not have been able to go through the Build an AI Agent experience. So...insert this session!

Join me and ServiceNow MVP @John Zhang1, tomorrow, Tuesday, May 27th at 11AM ET, to gain insight into the intricate details that make ServiceNow's AI Agents so powerful. Not only will we answer the question of how they're different from a chatbot or a well-built flow, but John will showcase two real-world examples of agentic AI (beyond what's included out of the box).

I look forward to seeing you there!

Please consider marking my reply as Helpful and/or Accept Solution, if applicable. Thanks!

The deck presented has been attached to the event (above).

In the next day or so, the recording will be posted to the ServiceNow Community YouTube channel and the Ask the Experts playlist.

Thanks!

Please consider marking my reply as Helpful and/or Accept Solution, if applicable. Thanks!

Recording is now available!

Please consider marking my reply as Helpful and/or Accept Solution, if applicable. Thanks!

Hi everyone,

Here's a list of the questions and answers from this Ask the Experts session:

Question	Answer
I'm not an expert or a python developer, but I keep hearing a lot about MCP servers. Will ServiceNow offer support for MCP servers in the future?	Hi, in this context, ServiceNow is the MCP Server (where AI Agents from any technology could pass communication with), but as far as an MCP Client, ServiceNow does have an offering for that: https://store.servicenow.com/store/app/5eeda18f1b996a10229141d1b24bcbfc
I might be optimistic but do you see sometime in near future there might be a separate role as an AI Orchestrator? We have BA, BPC and lots of non tech roles but I think AI Orchestrator will be a low code developer having good process sense?	Absolutely, there'll be roles such as prompt engineer, AI Agent engineer, etc. - today, developers, admins, builders, are the ones who are setting these up.
would the comparison of the workflow vs AI agents be same as comparison to sched job+AI?	Hi, yes! Scheduled job/flow are the same in this context.
So is control tower a manager of agents?	Yes! More info here: https://www.servicenow.com/products/ai-control-tower.html
Newbie Question: how to identify a use case if it qualifies to be implemented using AI agents VS. Automation/orchestration using ssh, powershell, APIs etc. (from the perspective of cost of implementation and maintenance)	Hi, I have a Leading Practices slide that we'll cover near the end that explains this! 🙂
AI agent discussions commonly focus on ITSM. What are the exciting use cases that would apply to ITOM, CMDB, Discovery?	There are hundreds of out of the box AI Agents across all of our products, a few callouts from ITOM/CMDB/Discovery are: - Triage and Analyze Alerts - OT Device Onboarding - Service Taxonomy Creation - Problem RCA (data collection and analysis) We also have additional AI Agents added quite often on our ServiceNow Store page.
What is the difference between AI Agents and Virtual Agent? Will AI Agents replace Virtual Agent? Does AI Agents use both LLM and NLP ?	On a previous slide, we discussed the difference between just chatbots/flows and AI Agents, we'll be sharing the deck after this session as well. As far as AI Agents using both LLM and NLP; AI Agents are task based/workflow based, not necessarily something you'll always chat with.
Are there any Labs like we have for OnDemand courses to learn / implement AI Agents?	Yes, there's several training courses on our Learning site and there'll be some posted from our Knowledge 2025 conference in a few weeks. A few available now are: - https://learning.servicenow.com/lxp/en/now-platform/now-assist-ai-agents?id=learning_path_prev&path_... - https://learning.servicenow.com/lxp/en/general/ai-agents-vs-traditional-ai?id=learning_course_prev&c... - https://learning.servicenow.com/lxp/en/now-intelligence/ai-agents-delivery-accreditation?id=learning...
As we continue to mature our Discovery and CMDB processes, what is the best way to begin incorporating AI agents into our workflows? Are there any recommended low-effort, high-impact use cases we can implement as a POC before scaling to more complex scenarios?	Within the ITOM/CMDB space there are some AI Agents that come out of box, but these initial use cases are focusing on the health of the CMDB. Think reviewing current CIs, monitoring the introduction of new, identifying trending issues or gaps in data and helping create a plan, referencing organization process and knowledge articles to craft that plan, etc.
AI Agents and Agentic AI same things?	an "AI Agent" is singular, "Agentic AI" is more so about the "framework" of using multiple AI Agents. So...similar, but one is singular or talking about them in a singular way and Agentic AI is the concept of an AI framework.
Do we need to install somethign to get this?	Hi, yes, you'd have to install AI Agent Studio.
Does AI Control Tower come with its own range of pre-built agents for Asset lifecycle, Risk and compliance and Case Mgmt that help you get all those analytics? I am bit confused on how AI Control Tower is different to the other Pro plus offerings	ServiceNow AI Control Tower is a centralized AI platform that helps organizations govern, manage, and optimize all their AI initiatives. It's designed for AI Centers of Excellence (AI CoEs) and Chief AI Officers (CAIOs) who need visibility and control to scale AI responsibly across their enterprise. More information can be found here: https://www.servicenow.com/products/ai-control-tower.html?state=seamless - feel free to comment below if you have any additional questions!
Will AI Orchestrator a separate role in future with combination of low code developer having good process sense?	In the future, there will most likely be new job roles within the ServiceNow ecosystem such as AI Prompt Engineer, AI Agent engineer, AI Agent Orchestrator (as a role), etc.
Could you please provide more information on how the NOW LLM works, its setup, and developed, particularly in the context of resolution notes generation or any other use case?	Here is a good resource to get acquainted with NowLLM: https://www.servicenow.com/community/now-assist-articles/now-llm-generative-llm-for-enterprise-ai-us... - additionally this ServiceNow YouTube video is quite helpful: https://youtu.be/fxfzkGO0_xo?si=-BCHJGHNcGDyQJKi
All these questions asked will be documented somewhere so that we can refer to them later along with the ppt deck?	Yes! They'll be attached to this session on our Ask the Experts Community page: https://devlink.sn/asktheexperts and the PDF version of the presentation will also be made available there too.
If the end objective/Goal requires multiple hops on different tables and data retrival then can we employ multiple agents ? like 1 lead, 3 engineers, 1 QA etc ?	Yes! This is leading practice to break complex tasks into multiple AI Agents (each focusing on their specific area of responsibility).
How does Now LLM determine the appropriate model—such as a large or small language model—for a given problem?	Within the AI configuration in an instance, you're able to specify what LLM should be used. Whether that's the NowLLM, Azure, or other external LLMs.
In the demo, I noticed the term AIA ReAct in the logs, which stands for Reasoning + Acting. Could you please provide more details on the framework ServiceNow is using?	The ReACT (Reasoning and Action) framework we use is GPT-4o.
If no trigger condition is specificed for a use case, does this mean the Orchestrator scan the AI Agent each time/performance intensive?	If there's no trigger condition, then that use case won't trigger any follow-on activity for the Orchestrator to query which AI Agent is best to use.
In the demo, I’m unable to see the total response time taken by the complete query during testing. instead, it is showing only for each step	The time for completion varies per model you're using and the AI Agent's configuration. Feel free to comment below if you're looking for a specific metric.
can we enable AI agent studio plugin in PDI ?	This is currently not available in PDIs
is there a way where we can practice/exteriment with AI agents ?	There is an FAQ regularly updated by our AI team located here: https://www.servicenow.com/community/now-assist-articles/ai-agents-faq-and-troubleshooting/ta-p/3200... and there was opportunity to build an AI Agent at Knowledge 2025. Throughout the rest of the year, there'll also be other opportunities to build an AI Agent at events such as CreatorCon India and the various World Forum locations.
Is it possible to integrate an ai agent to steer Virtual Agent conversations with end users?	Virtual Agent would use Now Assist for it's conversations, but in theory, you could use an AI Agent to be triggered by an interaction to then do something, but you have other mechanisms available as well beyond using AI Agents for that.
How does Now LLM determine the appropriate model—such as a large or small language model—for a given problem? For this question, This is the followup question, when creating a skill there is only NOWLLM and there is no NOWLLM large model or NOW small model for selection	There is only one selection for NowLLM as it's one model. When using the Now Assist Skill Kit (NASK), you can use whichever LLM you'd like. More info can be found in articles like this: https://www.servicenow.com/community/now-assist-articles/now-assist-skill-kit-knowledge-article-coac...
Are these AI Agent front end user facing? or it will be still chat bot or VA?	There is Virtual Agent with Now Assist. AI Agents could potentially be triggered by end-user activities, but Now Assist within VA gives you the ability to leverage LLM(s), uses Natural Language, AI Search, etc.
do various AI agents need to be set up first in order for the orchestrator to use them?	Use cases/AI workflows will have a "trigger" that should be configured. Once it's triggered, Orchestrator will review the AI Agents and decide upon the most appropriate to use. So yes, AI Agents need to be set up first, so that Orchestrator can see them and know which are available for use.
Is there a target release for the AI Control Tower?	This product was announced at Knowledge 2025 (May), with a potential release scheduled for July 2025. This can of course shift, but for now, that's the plan! Here's a link to a video we just released that covers the AI Control Tower in more detail: https://youtu.be/kC_uwr8Dk9o?si=zRAFWOX5_Stv0v15

Please consider marking my reply as Helpful and/or Accept Solution, if applicable. Thanks!

As of now, all questions from our webinar have been answered above. If you have follow-on questions or would like more information, simply let us know below.

Thanks! 💚

Please consider marking my reply as Helpful and/or Accept Solution, if applicable. Thanks!

Hi @Allen Andreas ,

How can I get more clarity on the comparison between the Flows/Scheduled Jobs with Gen AI prompt and new Agentic AI Workflows? As I'm bit less aware about first approach.

Thanks!

Join the Community

Let's connect together

Join us to make our over 890K member community even better. Create your free account and make the world work better for everyone.

Join the Community