If you missed the webinar “Driving business transformation with ServiceNow AI Agents” on March 26, 2025, we encourage you to watch the recording to review what our AI experts Ritesh Shah, Ashish Sethi and Tara Fisher discussed. In addition, here's a recap of the questions asked by the audience.
Q: Does it (ServiceNow AI Agents) support bringing your own agent?
A: This will be supported in the Q1 release with the addition of the AI Agent Studio which will allow you to create your own agents. It will be moderate at the initial launch to ensure we are co-developing and then should move towards full control by the second half of the year.
Q: Any user manuals available for developers?
A: There is a lot of content and links that provide info to developers. In additional there are AI academies hosted every other week via the Community where speakers get behind the scenes.
Q: Can we jump directly to a scenario of AI Agents without first using/implementing Now Assist for ITSM, ITOM, HRSD, etc.?
A: You do not have to use Now Assist and can use AI Agents autonomously, but you will find that the combination will get you to high level business goals quicker. Like using Agents with Now Assist Virtual Agent will get you towards full automation that much faster, or using Now Create skills to move that needle quicker as well.
Q: How do end users interact with AI agents? Existing Virtual Agent chat only or will there be a dedicated new “AI Agent UI”?
A: AI Agents will run behind the scenes so the actual end user will not know what is happening unless the agent is set up to engage with the user for a specific question. For now, we have the human in the loop so it may need to get approval at one step to move on to the next step.
Q: With Agentic AI being framed as no longer being about assisting, and more about taking actions, what is the plan for the licensing model? As my understanding is that this is based on consumption of “Assists”.
A: We recommend you discuss this with your Account Executive.
Q: What is the benefit of using AI Agents for the example with Software Request Fulfillment, compared to a regular flow?
A: It really depends on your flow, how many steps do you have – could some of these steps be eliminated and handled by an agent? The idea behind the agent is that it will learn and do the thinking and decision-making that would normally require human intervention. While a human will remain in the loop for priority business decisions, the agent owns the other steps such as checking licenses, determining if a user is entitled, etc.
Q: Very curious how you produced those ROI data points. Are those generated through a Now Assist dashboard of some kind? We’re struggling with tying the AI functionality in ServiceNow to some of the tangible values in that slide.
A: Part of it is coming through the Now Assist dashboard and part of it is being tracked by teams through reporting and Predictive Analytics.
Q: What are some of the common reasons that would prompt us to use our own LLM instead of that provided with ServiceNow?
A: While we believe our LLMs are the best for all ServiceNow standard use cases, some customers have spent time building their own LLM’s that may give them a competitive edge or are very specific for their industry.
Q: Would the knowledge graph also be available as a tool in this architecture or this an additional level/layer?
A: Currently, Knowledge graph is an additional layer; integration with Virtual Agent / knowledge graph is available on a limited basis right now.
Q: Does AI Agent Orchestrator allow: 1) Monitoring of performance of each Agent that makes up that process 2) Can you start building an AI Agent Orchestrator process using a Business Process Model? 3) Can a human agent be engaged if a process is not going as expected?
A: Orchestrator triggers the agents depending upon the task or specific work and it monitors their activity. If an agent doesn’t know the answer or the task is to be done by someone else it can pass it on to another agent. 2) Orchestrator is built in and you can pick a business process to solve for it, 3) Yes, command can be passed onto a human agent as needed.
Q: Other vendors are trying to play the role of AI Agent Orchestrator – for example Microsoft, Ui Path Maeastro). How can we (customers) differentiate when or where ServiceNow AI Agent Orchestrator is a better use case vs using another vendor orchestrator that can also call a ServiceNow Agent alone?
A: One of the things to keep in mind is that your company purchased ServiceNow to automate your workflows and improve your processes, ultimately to get as close to full automation as possible. AI Agents is just the next step in that journey, it’s the natural progression with our product. Unless you have to leave the platform, you want to stay in there when possible to maintain that single platform of record.
Q: Can a skill created by us using NASK be used by an AI Agent?
A: Yes.
