Making Sense of AI on the ServiceNow Platform

Christine Morris

Welcome, everyone, and thanks for joining us on another Platform Owner MasterClass.

Today we're tackling something I think just about every ServiceNow platform owner is struggling with right now — AI, AI, AI.

It seems like every time we get comfortable with one of the new ServiceNow products, they announce three more.

So we've got Now Assist. We've got Agentic AI, we've got Moveworks, Slate, Otto, and the list just keeps growing. And while all of this is exciting, it also creates some pretty legitimate questions. So what do these products actually do? How do they fit together? What do we really need to support them? And perhaps most importantly, how do we make sure we're solving real business problems instead of just buying more technology?

And that's what we're going to go through today. I am Christine Morris.

I am Vice President of Consulting and Strategy here at Ondaro.

I've been in the ServiceNow ecosystem for about 11 years now. I started as a platform owner and worked at a couple of different organizations. I started the Richmond SNUG — I loved it so much, the whole platform, that I changed my career. And all I do now is really focus on helping clients fall in love with the platform as much as I do.

Ron, want to introduce yourself?

 

Ron Burt

Sure. My name is Ron Burt. I'm the Associate Director here at Ondaro. I work with some amazing people like Christine, and I have 27 years in a higher ed environment before coming to Ondaro. I started about nine years ago with ServiceNow in Kingston. I first initially hated ServiceNow — I got dragged kicking and screaming into it — but once I got into it, I really started to love it, love all the relationships, loving having all the data and everything together and what you can do with that.

And now I eat, sleep and breathe ServiceNow and enjoy it. So looking forward to talking to you guys today about AI.

 

Christine Morris

So today's agenda — we're going to talk about the AI explosion, decoding the ServiceNow AI portfolio, talk about some AI in action use cases, the platform owner's new responsibilities.

As if we didn't have enough to be responsible for before this — how you can get started on building your AI roadmap, and we'll save some time at the end for questions.

Before we dive in, I want to talk a little bit about Ondaro and what we do here.

So we are the only pure-play ServiceNow partner with fully certified resources across the platform. You can see all of the capabilities there on the left. If it's in ServiceNow, we take care of it. And really that falls into kind of three pillars for us.

Our Envision pillar is where we're really doing business transformation. We've got a team that focuses on organizational change management, AI readiness, platform strategy and governance.

We've got our technology implementers — the folks that are architecting and engineering, doing product implementations. We've got an amazing award-winning UI/UX design team.

A very large app dev team that focuses on App Engine.

Of course, all the integrations you could think of.

And then in addition to that, we also help folks with just ongoing day-to-day support through our Propel offerings of platform and operations, product management, those types of things.

A little bit of housekeeping.

We want this to be interactive. Feel free to use the chat, ask questions.

We want to make it as interactive as possible.

So a little bit of history — this is our third MasterClass series.

If you missed out on any, we'll put the link in the chat and you can take a look at those. We did a lot of comprehensive sessions on CMDB. We still have our ongoing ITAM MasterClass. Our next one is going to be focused on where assets meet people automatically —

the employee asset lifecycle on November 17th.

And we've had three previous sessions. So if you've missed anything, please take a look.

Encourage you — lots of great content out there.

And if this is your 50th MasterClass session with us, you know we love our survey poll. We'd love to hear from folks — where is your organization on its ServiceNow AI journey?

You can put your responses in the chat. So A — exploring, not yet licensed; B — we're piloting Now Assist or a few skills.

C — we're running AI agents in production; or D — we're still sorting out the product names.

Garrett — exploring, not yet licensed.

Lots of A's.

Lisa — hello.

Piloting Now Assist.

Jerry — still sorting through product names. We're going to help you with that.

So it looks like everybody's really kind of falling into the just-getting-started aspect of it. So hopefully this session will be very helpful for you today.

I want to start with why this has become such a big conversation, and why it can be surprisingly difficult to figure out where to begin. I love this meme: "Who are we? CEOs. What do we want? AI. AI to do what? We don't know. When do we want it? Right now."

And I think all of us are feeling that pressure in our organizations. So let's start to decouple the mystery.

This is where I think organizations get themselves into trouble. Every executive is asking about AI. They want to know what we're doing with it, when we're implementing it, what kind of return they can expect.

But the expectations look very different depending on who you ask. Business leaders want faster service, lower cost, and proof that their investment is paying off. Employees don't necessarily care whether they're interacting with generative AI or an agent — they just want to get an answer, complete a task, and get on with their day.

And then there's the platform owner sitting in the middle of all this fun, trying to figure out what to implement, how to protect the organization, how to demonstrate value. And the mistake is starting with the technology instead of understanding those different needs.

I can give you the most sophisticated AI agent in the world, but if it doesn't solve a problem that somebody actually has —

then we haven't actually accomplished anything.

Let's start with our capabilities right — before we start throwing around ServiceNow product names, I want to simplify AI into the four things it can do. First, it can find information. Think about an employee asking where to find the travel policy.

Second, it can generate something — maybe summarize an incident, draft a knowledge article, recommend a response.

Third, it can act. Instead of telling someone how to reset a password, it can actually initiate that approved process. And fourth, it can orchestrate. That's where things get interesting — instead of completing one task, AI can coordinate multiple activities across systems and teams. Think about onboarding a new employee: there are HR activities, equipment, access request approvals, potentially facilities involvement.

And the more autonomy we can give AI, the more important our processes, our data, permissions and governance become. And that's the theme you're going to hear throughout today's session.

And now I'm going to turn it over to Ron to talk about decoding the ServiceNow AI portfolio.

 

Ron Burt

Awesome. Thank you, Christine. In this section, I'm really not going to go through a full walkthrough of all the feature catalogs and things. The goal of this section is for you to get a mental map so that when you hear those product names getting thrown around, you kind of know what layer they fit in the process.

Because as Christine mentioned, over the last 18 months, there's been a lot of stuff that ServiceNow has added — Now Assist, Moveworks, AI agents, Otto — you know, all these different things, the Control Tower. And so it's normal to feel a little left behind. So we're hoping that by the end of this section, you can kind of explain to your CIO, explain to folks in a sentence or two about what each piece does and how they all relate together.

So we're going to jump into the next slide on how do those pieces all fit together. There are really three layers, and they all work together, and you need to govern them from end to end. In the top section around employee and agent experience, this is where people are going to interact with AI.

Go down one layer below and you have that intelligence and execution layer — that's where AI does its reasoning and acting. And then at the bottom is your workflows and your enterprise data, and that's really what's going to make AI useful. And then over on the right we have this AI Control Tower that kind of governs all three of them.

And it runs alongside all of them. AI is only as good as the layer beneath it. If AI can't see HR cases and records and policies —

if it doesn't have access to those, then it's just a search bot with personality.

So you need to have those layers. If we start at the top layer — we've got Otto, Moveworks, Employee Slate, and Employee Center.

We're going to do a little deeper dive into each one of those in other slides. But right now, that's going to be the employee engagement experience layer. The middle layer — the intelligence and execution layer — is where you have Now Assist skills. Those are going to assist a person. As Christine mentioned, they're going to help you summarize, help you draft, suggest things for you.

The AI agents next to that — those are going to act towards the goal using some tools.

They're going to do a little more than just assisting. It'll assist you, but it'll do more acting towards the end product. And then you have your workflows — those are going to orchestrate several different agents to get a business outcome. And then you have your autonomous workforce, which is going to package those agents as role-based AI specialists.

So it really is that mid-layer of the execution and intelligence piece. And then the bottom layer is your workflows, your data. You have ITSM, CSM — that's where you have your information, that's where your data is being captured. And then you also have App Engine, which covers some custom processes and workflows.

Then your workflow data fabric and integrations — that's where it reaches out to data outside of ServiceNow. So if we use an analogy of a restaurant: the front of the house is that employee and agent experience that people come into. Then you move from that employee experience to the kitchen, which is where the intelligence and execution is happening and things are being made.

But if you go a little bit further, you go back into the pantry and the supply chains that are supplying the kitchen with the information — or the food — to deliver to your clients. And then over on the right-hand side with the AI Control Tower, that's really going to be kind of like your health inspector and restaurant manager all at once, looking at everything and making sure that everything's

running smoothly and not going off the rails and not going to poison anybody.

So that's kind of — just a minute. Yeah, go ahead.

 

Christine Morris

I was just going to add — and now that we just got used to the name AI Control Tower, ServiceNow has renamed it to SI Control Tower. And the SI stands for Super Intelligent. So you may see that as you're navigating out there.

 

Ron Burt

Right. So let's jump into the top layer — the public-facing part. We're going to talk first around Now Assist. And Now Assist is where most customers really first meet generative AI because it shows up where people are at, where the agents are already used.

There's no new portal, there's no new habit. From an IT example, a service desk agent opens an incident and there are 40 work notes and it's been reassigned three times. They can do this incident summarization right from the interface, right from Service Operations Workspace, right from where they're already working.

So they don't have to go somewhere else. And when they close that out, it's going to automatically give them some resolution notes based off of the activity stream and things that have gone on with the incident. So it kind of works in the environment where those agents are already used to working.

So it has — there are a couple of things. There's incident and case summarization. You have resolution note generation. And that's going to work on prompts —

it gives instructions and guardrails. So those are skills.

There's knowledge search and article drafting.

And you also have agent productivity and assisted decision-making.

So it can help your agents make decisions and be more productive based off of information it's finding. How the skills work though is — there's the workflow context. So that's the record, the related knowledge, the user and what role they have. So it kind of works through

workflow. The skills are going to be a packaged job with prompts, instructions, and it puts some guardrails around what they can do.

Then the model — the model can change, the skill can change. But the skill and the workflows, those stay. So it's like replacing an engine — you don't have to redesign the whole car. You can just take out that part and it's going to be able to still

be implemented with the new.

But that does have governance implications. When you do change a model out, you're going to have to retest it to make sure that the skill output is what you're expecting to see.

And then at the very bottom you'll see "person." This is a very key human-in-the-loop line. We want to make sure — because I know for you guys, you've used AI and never had it hallucinate or do something strange for you in the world —

so you always want to make sure there's a human in the loop that makes sure nothing is put out there without some validation or verification. And a point to ponder: where are your people spending time reading or evaluating before they act? If you think about that as you're going through your processes, those can be some good candidates for places where you can start to look for having Now Assist

to help you out with those — to help guide the time they spend and reduce the time spent reading

or evaluating.

 

Christine Morris

Just going to jump in on Now Assist — it is only as good as what's in your platform. You have to have good knowledge. If your folks just close an incident and don't put the details behind it, it's only as good as the data it has access to. And so that's one of the things that's so important — one of the reasons

we do AI readiness assessments is because we've seen quite a few customers just turn on Now Assist and then wonder why they're not getting the value. You've got to have the right data on the back end, the right processes on the back end, and be consistent in that to get the value from it.

 

Ron Burt

And so let's talk about AI now — and this is going to move from recommending to acting. In the traditional sense

of traditional automation, it's going to follow some predefined rules. A workflow is designed as it's designed, and it's going to be assigned by category. So let's say somebody reports VPN won't connect.

Traditional automation is going to say, well, this is a network thing. It has a category of network, so it's going to assign that over to the network team. And that's where it kind of ends right there. That's traditional — assigned by category.

If we move over to generative AI —

this is where you get a summary, maybe a suggested fix.

And then it's up to the agent to decide whether they want to apply that or not. So it gives them some content recommendations, a person is going to review the output, and then they're going to — for example, like draft resolution notes. As Christine was mentioning, the more data you have, the better those resolution notes are going to be.

Especially as you start tying things together — with your CMDB and other things, you can start to gather really good details about what happened, especially on the incident management side. But it's going to draft some of those, and then it's up to that agent to push it forward.

On Agentic AI, this is where it's going to actually go through and take action towards the goal. The agent is going to do that within a set amount of boundaries.

For example, with that VPN — it's going to diagnose and fix. So it could go in there, find an outdated client or an expired certificate based off of the information provided. That could trigger a remediation, it could confirm the connection works, and then document what it did and close it — or escalate it.

So it could actually do all of that together. Again though — generative AI mistakes are drafts that you can go in there and tweak,

but Agentic AI mistakes can have actions that you have to undo. And that's why there is a risk.

And again, you want to make sure that the guardrails are set up and the operating model and governance are all set up.

So that it's not making those changes unintentionally — you want to make sure there's clear governance around that.

What makes it work? Again, you want to have clear process ownership. Back to that VPN — who owns the VPN process? You want to have reliable data.

Is the information right about this user and device? Do you have the right agent permissions — what does that agent run as and what can it touch? The human approval aspect — where in those steps do you need a human to intervene and approve? Monitoring — who reviews what that agent did and makes sure it's being monitored.

And then if something gets stuck or there's a problem, what's that escalation path?

So really — who approves? What are the boundaries? And what happens if there's an escalation? That's what makes this work by answering those questions. And then the building blocks you want to understand are those AI agents —

those are goal-driven workers. You're going to have the workflows which coordinate several agents — flows, scripts, API calls allow those actions to be taken. Orchestration decides who does what, where the human approvals are, and what those boundaries and limits are

that are going to be set within those. And then you have the autonomous workforce, which is ServiceNow's packaged agents as role-based AI specialists.

So they kind of build on each other. And those are some of the building blocks that you'll want to understand as you move through

this Agentic AI.

This is the experience layer. This is the slide that you want to take back to leadership.

And this is the big one. There are four different —

these are not interchangeable chatbots, okay. Otto is a unified AI experience. It combines the Now Assist and Moveworks AI experiences, and it was announced at Knowledge 2026.

And it's bringing those together.

It appears first in Employee Works and the AI Control Tower, and then it rolls out across all the products.

So think of this as like one assistant that employees will recognize.

The next one down is Moveworks. And this is that conversational engine and enterprise search. Its strength is that it reaches across all applications, not only ServiceNow. So employees can ask and get answers or actions from whatever system holds it.

The power with Moveworks and Slate — Moveworks remains available as a standalone, so

there are still standalone offerings if you want to keep those as well. Slate is that AI-first destination for services, tasks, information and interactions. So

it's kind of a personalized view that an employee can open. And they can see that they have two approvals waiting,

a laptop refresh due, an open HR case, and a box where they can simply ask for whatever they need. You can have all of this in a single pane of glass — all that information together.

And then there is still Employee Center and Employee Center Pro — that's the established portal that's out there.

And we just want to throw that out there that it's not being thrown away.

Slate is going to inherit some of the configuration — inbox profiles, charts and all that kind of stuff. So looking at it from an analogy: Otto is going to be kind of the storefront.

Moveworks is that engine inside. Slate is the new store layout, but Employee Center is still there — it's an existing building being renovated, not demolished. So any work or effort you're putting towards Employee Center won't be

for nothing as you move forward with Slate as well. Your investment is protected.

 

Christine Morris

And one thing I'll mention — Slate doesn't have all the bells and whistles that Employee Center Pro has yet. Some of the news and announcements and some of the things that really drive the employee experience are coming later on the roadmap. And what we're seeing some clients do is they want to take advantage of Moveworks and Slate, but

if you've made a significant investment in your Employee Center Pro, or you're using it for an intranet, sharing news and giving people the ability to search and browse,

you can also surface the Moveworks bot right there. So you don't necessarily have to convert to Slate to utilize those features.

 

Ron Burt

Good point.

And here are the two screens. On the left you have Employee Center Pro —

that's the model portal you have: the taxonomy, your catalog, your search and all that. And then on the right you have Slate, which is more of that conversational-first, persona-based interface with dynamic entry points that can be highly personalized.

From an example perspective — a new hire needs a laptop, a parking badge, and benefits enrollment. In Employee Center Pro, you go to three different places to get things done, or you can do an order guide, but you're going to be in

different places doing different things.

Whereas with Slate, you can say, "I start on Monday, what do I need to do?" And a whole checklist shows up and you just go through it. They don't have to navigate to work — they just get work done.

It goes through that checklist and does it all right there in one place. But again, Slate is going to build off of existing Employee Center configuration. So it's an evolution, not a rip and replace. Just keep that in mind.

 

Christine Morris

Yeah. And the other thing to mention about the Slate and Moveworks piece is it starts to interact with other systems. So if you connect to Concur through Moveworks, you can enter an expense report right from that Slate screen. You can attach a receipt and it's going to go take action in Concur and do that.

And that will segue nicely into Ron's next slide, which is about the workflow data fabric.

 

Ron Burt

That's right. Because as Christine mentioned, most of the data that AI needs or agents need lives outside of ServiceNow — or at least a lot of it does. You have HR records, ERP data warehouses, document libraries. And so workflow data fabric is how the platform is going to reach out to those different areas.

The first one down there is query in place. This is zero-copy — where a customer's agent sees a customer's recent order from a data warehouse without having to do a nightly import into ServiceNow. So you're not waiting until a certain time when all those

jobs run.

You're actually getting that live — able to query it in place.

The second one is integrated stream. Integration Hub spokes connect to those third-party systems without any code. And Stream Connect can bring in real-time events.

So for example, a lost badge event from a security system can automatically start a workflow.

You can have those integrations. And then search content —

this is where external content connectors let people search SharePoint or Confluence

from ServiceNow while still respecting those source permissions. So finance-visible documents stay visible to only the finance folks. You can still have that

autonomy — people can see what they're supposed to see and not see what they're not supposed to see. And then added business context: the Knowledge Graph and data catalog that relates data to people, assets and processes. So questions like

who can approve Maria's software request can be answered because the system knows who Maria's manager is, what the cost centers are associated with that,

and it knows the laptop. Having all that business context using Knowledge Graph and data catalog to gather all that.

And in the next few slides, I'm going to show an actual example of where an agent handles a relocation process.

But a key takeaway from that one is that if you have fewer copies of tables to sync, store and secure —

you still have to manage the data.

That's what we want to make sure of — you still have those data contracts.

You're still the owner of the contracts, the source owners. And you want to make sure that you know who owns the source, how fresh it is, and what those agents see. From a platform owner perspective, you want to make sure that you're still

on top of those contracts and things of that nature.

So SI Control Tower — building trust at scale. This was released at Knowledge 2026.

They expanded SI Control Tower. And the goal of it is to discover, observe, govern, secure, and measure AI across the enterprise. We want to make sure

that even AI that's not built on ServiceNow is covered.

So there are a couple of things we're going to walk through here.

On the discover side, this is going to find agents — maybe an agent a business team built on another platform that nobody registered. Shadow AI becomes visible at this point. So it's going to discover

different AI or shadow AI that's out there.

Then it's going to observe. It can monitor agent behavior. So if an agent suddenly attempts a couple hundred password resets in an hour — okay, that's unusual behavior. Let's dive in and see what's going on. So it can observe agent behavior as it's happening.

And then on the governance side, it can assess risk.

So it can go and do risk assessments before an agent touches payroll data before it goes live, or things like that. So it allows you to assess the risk against agents, model data sets and prompts.

Secure — it's really identifying and scoping the agent's identity and permissions. And it can stop an agent that

tries to overreach —

it's going to be able to stop that agent from overreaching. And then measure —

it's going to track cost, adoption and your ROI, your return on investment.

For example, every cost of a resolved case before and then after. So think of the SI Control Tower as kind of a DBA for AI combined with an air traffic controller. It's keeping the guides on it

but allowing it to do what it needs to do.

And it is a governance product, not a governance program. So the Control Tower can tell your agent that it has no owner, but it can't decide who the owner should be. So owners, decision rights, security policies, accountability — all of that still sits with you

to make those determinations.

That's kind of the portfolio of it. Now we're going to see it in action with one real person.

In the next slide, we're going to see how does that work with one employee, one request, and a better experience. So everybody say hello to Sally. She's a product manager in Chicago who's moving to Berlin next month. The today version first — right now she opens an HR case, raises a separate IT ticket for Europe with power adapters and keyboards, emails facilities for desk and badge, hunts for relocation policy about payroll and tax changes, and then spends weeks chasing five different teams around for updates on all those different requests.

Now let's look at what's that better experience.

She goes in there and asks in plain English what she needs.

It's going to retrieve and pull that relocation policy, her employee record, wherever those live.

So she's going to go through the context of it.

Then you're going to go into the planning, which is the AI and agents.

That's going to cover the HR, payroll, the IT and facility tasks. So it's going to break those down into the HR transfer, IT equipment request, facilities desk, payroll changes — all of those.

And it's going to go into the acting — opening those requests and routing them for approvals. And that's where it goes into the execution side, into the workflows. Some of that can reach into HRSD, some into IT.

And then ultimately you get into the governance.

And this is where we're looking at that SI Control Tower — logging what each of those agents did, keeping compensation data limited to only the authorized people who need to see it. So as you're talking about payroll and things like that, that information stays protected and it's making sure those permissions are in place.

So it's kind of again building that trust and measuring it, and

limiting who can see the approvals for that. And so what Sally actually sees is one conversation, one status view throughout, with some proactive nudges along the way —

like if documents for work permits are due, she's getting those notifications along the way.

But it really is that once, seamless conversation that she's going through. It's not a single AI feature — it's

that coordination across the various departments that used to have to be tracked down and bird-dogged around. Now it's all together into one.

And so a point to ponder — as you think about internally in your world, what is your "Berlin request"? What is something where somebody has to go through multiple things to get there that could be streamlined?

This is just a way to start thinking about what is something that you could bring into this.

How can we do that in a way that's easy for people?

Which brings up a good point, which is the next slide — and this is around how much autonomy should we give the bots? How much do we give? Do we surrender to them?

And so we have Sally's case right there. Let's use her case as an example.

There's this use case of summarizing an HR case.

You have a generative AI assistant. They're going to review it and edit it before using it. The human is reviewing it and not strictly just relying on what the bot said.

There's an answer to a routine policy question — the grounded conversational AI —

do I get relocation allowance, something like that.

That might be a sample.

HR is going to be able to own that content. If the policy is out of date, keep that in mind — AI is going to confidently give them a wrong answer today. So you really need to have good

governance and control around the content so that — garbage in, garbage out is what's going to happen.

So you need to make sure you have governance around that. And then executing a multi-step employee transfer with Agentic AI workflows — you're going to have approvals along the way: manager HR approvals, HR approvals for compensation.

A person handles those exceptions — like a visa delay or something like that — they can be in there to handle those exceptions as a human.

So the humans are approving the steps and then handling things that fall out as exceptions. For factors that could shift the decision — risk, obviously. What's the cost if the AI gets it wrong?

You want to ask yourself these questions.

Repeatability — how often does the same pattern appear over and over?

Process maturity — is the process documented and stable?

As you're thinking about some processes you're doing — is it a mature process? Think of it like change management. You've got your standard changes — is this a wash, rinse, repeat type thing you've done over and over again?

Then why do we need to build some more complexity? Maybe we can do that and give some of that autonomy. But again, it's about how mature your process is, the documentation, and then data sensitivity —

does this touch compensation, health, or PII? What type of data? What's the sensitivity? Keep that in mind as well.

So those are kind of four factors to be thinking about as you go through this

to determine — a password reset could be low risk, highly repeatable, mature, and low sensitivity. So it might be a good one for automation.

Whereas termination of an employee — that could be a little high risk, rare and sensitive.

So that might not be one you want to go with just yet. So yeah.

 

Christine Morris

Like, what's my salary? Right?

 

Ron Burt

Yeah, yeah. Exactly.

 

Christine Morris

Lots of sensitive data that you've got to kind of factor into that.

Absolutely.

 

Ron Burt

All right.

 

Christine Morris

I'm not sure if you saw the chat — everybody is loving your analogies.

 

Ron Burt

Oh, good. I haven't looked at the chat, but thank you. Awesome.

 

Christine Morris

I'm going to take it over from here and talk about the platform owner's new responsibilities — as if we didn't have enough to do already.

You know, if you think about everything as is, the platform owner is already responsible for — you guys like my octopus lady that I did use AI to create? — platform health, upgrades, integrations, security, governance, technical debt, enhancements, a constant stream of competing business priorities.

And now we're adding AI. But AI isn't just another application or feature that you're turning on. We're introducing technology that can generate information, make recommendations, and increasingly take action on behalf of employees and the business. So that means somebody has to make decisions about which use cases we pursue, who owns the agents, what data and models they can use, how we monitor them, and how we know they're delivering value. And then let's not forget adoption.

Just because we enable the capability doesn't mean people will use it or trust it, or that it will be accurate. As I said, it's only as good as the data behind it.

I don't believe that a platform owner needs to personally own every one of those decisions, because you're going to be integrating with other systems,

working with other teams to replace the things they used to do. But I do believe that the platform owner needs a seat at the table because these decisions will directly affect the health, security, architecture and long-term strategy of your platform. And if your organization hasn't figured out who owns these decisions yet, you're certainly not alone.

I think we're all struggling together with that.

With that, let's take the next poll. So the question is: who owns AI decisions on your ServiceNow platform today? We'd love you to respond in the chat. A — the platform team.

B — we have a central AI office or center of excellence.

C — each department decides on their own; or D — no one clearly.

Ace. We got a "B-ish" from Ben.

Like, this is B.

Jerry says A.

Yeah, it's definitely falling on our plates. Oh, poor Lisa says no one clearly owns it yet. And that's the scary one, right? Because then that's the worry that we all face. Especially now that you've got a million bots out there — Claude, ChatGPT, and so many of them are accessible by the browser.

Who is doing what? So it's definitely a challenge as we grow.

All right. Well, let's talk now a little bit about all the AI tools. One of the things I've said a few times is that this actually reminds me a lot of the early days of cloud computing. When cloud first took off, there were providers everywhere. Everybody was promising to transform

how we delivered technology. Everyone was excited about the possibilities, and organizations were moving quickly to take advantage of them. But governance, security standards, and enterprise architecture weren't necessarily keeping pace. We had different providers, different approaches to security, challenges integrating everything, and a whole new model for managing costs. And we then had to spend years maturing our cloud strategies, establishing governance and figuring out how to manage it all effectively.

So I think we're watching a very similar situation unfold with AI, except this time the technology is evolving even faster. We've got ServiceNow, Microsoft, Google, and countless other providers introducing AI agents with overlapping capabilities. While organizations are understandably eager to take advantage of all of this innovation, many are still trying to figure out who's responsible for it and how to govern it.

How are these technologies going to work together? And a big one — especially in the consumption model for some of these bots — how are we going to manage the cost? We've heard those horror stories where somebody goes out and implements a bunch of AI bots, and the next thing they know, they've got hundreds of thousands of dollars in cloud usage costs.

Figuring that all out is going to be a challenge. The difference is that AI isn't just another place to host applications or store data. We're now talking about technology that can make decisions and take actions on our behalf. And that really raises the stakes considerably, and frankly, can be a little bit scary.

I'm not suggesting that we need to standardize on one AI provider. Just like cloud, there will be legitimate reasons to use multiple technologies. But we should take what we've learned from cloud adoption and apply it now, rather than waiting until we've created an environment that's difficult and expensive to govern. Because the last thing we need is to spend the next decade untangling our AI investments the way we spent years maturing our cloud environments.

So this is where the conversation gets kind of practical. If we're going to let AI agents take actions, we need to think about them a little differently than traditional software. First — permissions. An agent should only have access to the systems and information it actually needs.

Second — sensitive data. We need to understand what information it's using,

where that information is processed.

And what our policies are allowing. We need to protect against manipulated instructions — just because an AI agent encounters an instruction doesn't mean that instruction should always be trusted.

We think about things like approvals, testing, auditing, and the ability to intervene if something goes wrong.

And then I want to spend a moment on that last item — model choice.

This has become a real operational consideration for ServiceNow.

When this first came out, it was the ServiceNow model.

As of the summer, they're now allowing different default model providers for certain out-of-box AI skills and agents, and they began moving toward third-party providers.

Should you care?

Let's say your security team approved a particular AI skill because they understood which model was processing the data. Now a Store application updates and changes the default model provider for that skill. You could potentially have a different provider processing the information, different output behavior, or a skill that stops functioning because the new provider isn't allowed under your organization's policy.

The important distinction is that deliberately configured provider settings are preserved under the described rollout — it's the defaults that may change. So model governance becomes a huge part of your regular platform operations.

Before applying updates, you want to make sure that you're understanding which providers are approved.

Afterward, verify the assignments and retest the critical skills.

This is a perfect example of why AI governance can't just be a policy document that somebody sets and forgets about. It has to become part of how we operate the platform.

And again — we've hit on this a few times, but I just can't emphasize it enough. I had to have a whole slide for it. Your AI is only as ready as your platform. It is not going to compensate for foundational weaknesses — it is going to expose them. So today, if your platform has poor knowledge quality, AI will deliver unreliable answers.

If you have inconsistent processes — some people just close tickets with no data, no information — the automation is going to break. If you've got weak integrations, the agents cannot act. If you've got unclear ownership, you're going to end up with governance gaps.

If you've got poor adoption planning, you're going to end up with value that never shows up.

Rolling out AI skills and AI search and all those things — it's just like rolling out a new portal. You've got to market it and let people know,

let people know that this is now out there and they can get things done faster.

In many instances, we create all these skills and don't go through a campaign to educate and let everyone know. And then you end up with people building their own skills or building

other agents that aren't as effective, and then you don't get that value.

Another survey. So how confident are you in implementing AI tools on ServiceNow? A — we have a plan set in place and we know what needs to be done.

B — we have a plan and some ideas but have trouble launching.

C — we don't have much of a plan, but we do have an idea of what we want to do; or D — at this point we need AI to tell us how to implement AI.

My personal favorite. Love to see you respond in the chat. Jerry says D — let's ask the bots. We've got some C's, and I think a lot of us are really trying to figure out what we can do. At least one person says a little from column A, a little from column B.

And that's the struggle — figuring out what to do. And that's why I always say the technology is the easy part. It's the people that's the hard part. Evaluating what are going to be the best use cases for you to start — where do you start?

And with that, we will talk about building your AI roadmap.

 

Ron Burt

Yeah. So let's talk about something that you can act on this week. We're going to go pretty fast because we're running out of time here. But start with the outcome, not the AI product. I want to define what readiness is real quick — and I think Christine kind of alluded to this.

Clean data, documented process, named owner, licensing, and the right controls in place. Making sure that you have this readiness here —

is kind of defining that.

So — high value, high readiness: prioritize for early implementation. Start with incident summarization — fast win, visible value.

You can go to "prepare first" around high value but low readiness — where you can start to improve your data.

If HR sits in three different systems, fix that foundation first. Be selective —

don't turn on AI for low-volume requests just because the feature is available. That is a trap.

Deploying AI just because it's easy — you want to make sure you're going to be able to show business value. If it doesn't hit the business value, be selective in what you turn on.

And on the defer side, this is for a process that runs twice a year and has no owner — really no readiness or business value.

Let's go to the next slide real quick. Because right here is key — if AI is working, what actually gets better? Making sure that 10,000 AI interactions mean little if employees are still struggling to get help.

It's not really the activity metric. You want to focus on making sure that —

as you start to implement the AI stuff, 10,000 interactions is a baseline. You want to be able to talk about

from the employee experience perspective: the effort, the percent resolved on first contact — without having to bounce it around and get multiple resolutions.

The satisfaction level.

From a service performance perspective — your mean time to resolution, the quality of the resolution notes, the number of handoffs and reassignments, team capabilities. These are things you can baseline — big lean Six Sigma style.

First you define what you're going to do, then you measure. Measure where you're currently at so that as you implement these things, you can go back against that baseline and actually

show the before and after. Now let's look at the last slide, which I think is the best slide here — what can you do over the next 90 days?

You can move from just an interest in AI to actually getting results. The first 30 days is around discovery — inventory AI that's already in use. And usually you'll find that there's more out there

than leadership

thinks there is. Look for

Now Assist features that are licensed but switched off, AI features that are in other vendors' tools, public chatbots that people are using on their own —

things of that nature. Interview employees and process owners — ask what wastes your time, where would you like to see AI help you.

Then baseline those service experience and readiness metrics. That's the discovery portion.

The next 31 to 60 days is prioritize. This is where you're going to look back at that chart and really select those high-value pain points — pick one or two employee pain points that can really drive some value that you could

fix and automate with AI.

Next, name the owner, controls and success measures before building — as we were talking about, define it before you build.

And then prepare communications early and address the whole "is this going to replace me?" question that everybody seems to have.

Make sure that you're communicating with those pilot teams early as you start to implement.

And then in the last part — proof, 60 to 90 days — pilot with some real users. One service desk team for four weeks or something like that. Get their feedback and measure the quality, time and employee effort.

And again, you're measuring that against the baseline. This is what we were at, and now what's the benefit we're getting from this.

Then refine, train, and ultimately publish that roadmap for implementation.

Keep in mind through all three of these you're going to thread people, ownership and measurable outcomes throughout every step.

So you can lead, as a platform owner, meaningful AI adoption without chasing every new product name.

You don't have to keep going after them all. You can take a measured approach. And then the last slide —

own the outcomes, not just the technology.

Start with the human problem and then choose the capability. Sally and moving offices came first.

The products that needed to be updated came second. Build trust and readiness and accountability from day one.

A governance product is not a governance program. Make sure you're building that trust, readiness and accountability. And then measure your successes by those business outcomes, not by the number of agents deployed or processes automated — measure it by whether you're meeting or exceeding those business outcomes you're trying

to address.

And you don't have to become an AI expert overnight. As Christine was talking about, things are coming fast and furious.

You're just really looking at how you can start to turn some of that potential into measurable business value.

 

Christine Morris

The other thing I'll add — from a ServiceNow perspective, we all know it's not the cheapest tool. And some of these AI components are expensive. So if you just dive right in and start building agents, a year down the road, somebody in finance or in leadership is going to ask whether you're getting the value.

That's why it's so important when you're beginning to establish the baseline, as Ron said, and a way to measure it so that you can report it. And the SI Control Tower is pretty awesome at doing some of those components and bringing it kind of that one-stop shop.

Ron and I would love to talk all day about AI, but we're about out of time.

Any questions for us in the chat? Let's see.

Did folks find this helpful? Do you feel like you learned something today? Do you have a better understanding of all the pieces and parts?

If you need more help — we do have offerings around this. We've got AI readiness and adoption. We've got some low-cost offerings we can showcase — what the SI Control Tower does, how to get started.

If you need help developing your AI strategy and roadmap, we can help you with that as well.

I can't say this enough — because I've lived through some of these nightmares of technology in my many years in technology — you have to have a governance mechanism, you have to have some way to control it. And that really is a huge component of what the SI Control Tower does.

We'll try to keep you guys current on name changes — ServiceNow changes them. But

I think you guys get the gist.

Thank you so much for joining us today. We love these sessions. If they let us, Ron and I would talk to you guys every day, all day about what to do.

If you're interested, reach out.

You'll get a survey.

We'd love to see how we can help you.

Thank you, everyone.

 

Ron Burt

Thanks to everyone.