Applying ITAM
Principles to AI with
AI Control Tower

Christine Morris

I'm Christine Morris, and I'm going to be here with Ian today to lead the session. Today we're going to do a little welcome and intro to tell you a little bit about Ondaro.

Then I'm going to turn it over to Ian to talk about AI as an asset and how we govern that as practitioners. We're going to talk about managing your AI assets using ServiceNow AI Control Tower, where do you start, what else AI Control Tower is used for, and then we'll do our next steps and Q&A.

A little bit about us as an organization. Ondaro — we are the only pure-play ServiceNow partner with fully certified resources across every aspect of the platform. You can see all of the products there on the right, and we do that really through these three key pillars. Our Envision pillar covers things like helping you through business transformation, AI readiness — a good topic for today — platform strategy and governance, how do you get the most value out of the platform, roadmapping exercises, things like that. And then of course we Implement and Develop. We've got some amazing architects throughout our organization, UX/UI designers, and we do a lot of custom app development on App Engine. In addition to that, we also have a leg of the organization that manages and helps clients to optimize.

We all know that budgets aren't unlimited and sometimes you need help. So these are products like Ondaro Reserve.

I am Vice President of Consulting here at Ondaro. I've been in the ServiceNow ecosystem — I think I'm up to about 11 years at this point. I was a platform owner myself. I don't think there's a job in the ecosystem I haven't done. I started the Richmond SNUG and served as the leader of that for several years.

And I'll let Ian introduce himself.

Ian Cahall

Hi. Ian Cahall, Associate Director and Principal Architect for Ondaro's IT practice. I've been working in and around ServiceNow for a little over a decade as well. Looking forward to speaking with you all today.

Christine Morris

A little housekeeping, as always. We want these to be interactive. If you've got questions, don't hesitate to put them in the chat. If you want to share something, you can come off mute. We love reactions — if you see content that you love, give us a big heart. And if you need to see captions, click on "More" at the bottom of Teams and you can pull in captions as well.

A little look at where we've come from for our MasterClass webinars. We've got our CMDB MasterClass with lots of great content there — if that's something you're passionate about, you can click on the link in the chat and check that out. This is our eighth ITAM MasterClass. If you're interested in seeing some of the others or you've missed those, you can find those as well.

And then a relatively new MasterClass we have is our Platform Owner MasterClass, where we're bringing platform owners together across various client organizations and talking about things that are meaningful to those folks. Check those out when you have some time — lots of great recordings and resources. On our last Platform Owner MasterClass, we did building a business case, which is very helpful.

If you need AI Control Tower, you can check that out and we can help you get started.

If this isn't your first one, you know we love our polls. We'd love to hear from you — just put your response in the chat. What's your organization's current AI strategy? A: we have a strategy and governance in place. B: every department seems to be figuring it out on their own. C: we're pretty sure someone is in charge, we just don't know who. Or D: we don't use AI at all.

Kudos — we've got quite a few folks who have already started to think about governance. A mix of A's and B's. Great, awesome.

A little bit of a mix, lots of A's and B's together too. Let's jump in and I'm going to turn it over to Ian to start talking about how we manage these assets.

Ian Cahall

Thanks so much, Christine. We're going to get started talking about AI as an asset.

The idea behind this first section is to help folks wrap their heads around why AI is an asset. For some people this is probably pretty obvious, but for others it may seem like we're in a new, totally different category and we should be thinking about it differently.

The goal of the next several minutes is to explain why it really is largely the same thing, and why we can apply a lot of the same principles from our previously existing ITAM practices to AI. For those of you that have attended previous MasterClass sessions, it's a really good thought exercise as we go through this to think about some of the previous sessions and the things that we've talked about, and how similar it will feel to what we're going to get into here for AI today.

A quick refresher — what is IT Asset Management? Very simply put, this is the process of managing the full lifecycle of your technology assets and their associated costs. We do that because we've got a number of different goals with asset management. We want to know what we own, we want to know how much it costs us to own and maintain those things, we want to make sure that we're getting the most out of our assets, and along the way, making things more efficient for the people that use them. By its very definition, AI is a technology asset. It is a piece of technology that we use. We're investing, in many cases, heavy capital into the use of this technology.

And it has a lot of overlaps with a number of other sections, which we'll talk about more. But as we get further into the segment, the things you should be thinking about are who is responsible for overseeing your AI adoption and usage? Based on your responses, it seems like a fair number of the folks in attendance today have somebody playing that role, but if not, it's certainly something to start thinking about.

What does AI really cost your organization? And importantly, how are you optimizing those costs? And then finally, who is paying for these things? If you've got a company budget for this technology, is it coming out of that? Are we doing back billing, chargebacks, etc.? The thing I'll really call your attention to is on the right-hand side of the screen — we've got the traditional IT asset management lifecycle where we're talking about planning, acquiring, deploying, operating, maintaining, and retiring our assets.

AI is going to play right within this space, although it does have a slightly different lifecycle that we will talk about in just a moment.

Artificial intelligence as an asset — it really is an asset just like your hardware, just like your software, just like your cloud resources. AI has a defined lifecycle. It has a defined data model that can be managed nearly identically to your software assets, and we'll talk a lot more about that here in a couple of minutes. Your AI costs can be optimized by analyzing how and when your AI is being used.

And shadow AI is just as prevalent — if not more prevalent now — than any other historical shadow IT your organization may have experienced. And that has to be addressed, especially because of how expensive artificial intelligence can be and the way its consumption model generally works.

Just like we talked about before, the AI asset lifecycle exists in parallel to that standard asset lifecycle. And I think it's really important to understand how many of these stops along this path match or mirror or are related to that classic IT asset management lifecycle. The first step in any asset's life is deciding that we're going to bring that AI technology into our environment.

The Onboard phase of the AI asset lifecycle is where we're being introduced to the AI technology and we're really defining the key fundamental data points about that piece of technology. This looks like a few different things, and we're going to talk more about the data model next. But this is where we're really wrapping our heads around the essential elements of what the AI we're considering bringing into our environment looks like.

Who's the model provider? What are our token costs? What's the context window? All of these things are really important. This is where we get the ability to define that upfront in our process as we then work through the rest of the lifecycle. And it becomes really important right away as we talk about the next stage, which is Assess.

This is where we're reviewing that AI for performance, impact, and compliance. How does this piece of technology fit within our organization? How does it fit within our industry? Is this going to fulfill the needs that we have for it specifically, or are we taking a one-size-fits-all approach? In which case we've got other things we need to contend with.

Once we go through this assessment — and this should be a structured thing — it's really important to note that if we're talking about assessing AI that's coming into the environment, we should have very specific things that we're looking to understand. This could be fulfilled by an existing TPRM process that you have within your organization, where you are doing a risk assessment for the model provider or the technology itself.

This could be a number of other internal assessments where we're considering our own internal compliance, considering the costs and what we need to do to be successful with that. But it should be something that's formal and structured — not a rubber stamp that says "yeah, we looked at it, we're happy with it." From there, once we've assessed it and assuming we have given this AI technology the green light and we're going to bring it into our environment, we need to go through the Build and Test phase. This is where we are developing the AI technology to work specifically within our environment, or for the use case that it's being built for. And this is a little bit different than a number of other types of assets that we would bring into our environment. Generally when we bring in software, it's largely purpose built.

Now, there are some exceptions to that, and ServiceNow is perhaps one of them, where we have to do some configurations and make it fit within our environment. But AI as a rule is much more amorphous — there are a number of different things that general AI technology can do, so we need to make sure that it will work specifically for our organization, our use case, etc.

As we do that, we're going to do a series of build and test cycles where we are performing prompt engineering, performing integration testing, all of those sorts of things, and then we're testing the outputs to make sure that we're getting consistent results that we like.

Then we can promote it into a Deployment stage. That deployment stage is where we are rolling out this technology to be utilized by the organization. Very simply, this is our active stage of the lifecycle. When we think about how this compares with the classic ITAM lifecycle, this is an analog to our deploy and maintain phases. Once AI is out there, we are keeping an eye on how it's working — are people running into issues, is it producing the results that we want, and we're trying to make sure that we understand all of the other qualifying criteria. What's our spend? How can we optimize that? How can we bring some of these things together?

And then of course we've got our Offboarding stage. It is really important to understand that at any stage throughout this asset lifecycle for AI, we can actually jump to the Offboarding stage. For many organizations, you may onboard and assess an AI technology, decide it's not for you, and then go straight to offboarding. It's still important to track those decisions and where those things have happened, especially if your systems are at all being touched by this AI through the Assess or Build and Test process. That way you've got a strong understanding of what's happening with the technology, and why and how it may impact the rest of your environment.

Next, we'll talk about the AI asset data model, and I think this is again something that is very relevant. Obviously, if you're trying to manage this AI technology, you need to be able to understand roughly how the components of our data are going to work. One of the things I'll call out here is that this is relatively similar to the software asset data model.

For those of you that have been in our previous MasterClass sessions, or maybe you're familiar with how SAM Pro works, you'll be familiar with this concept of having a model and then having other assets that are child records of that model. This is similar, but not exactly analog, and I'll explain a little bit of that.

At the top of our hierarchy, we've got AI Systems in the AI asset data model. An AI System is the parent record for all of your other AI asset data. And unlike some of your software models, an AI System model is a set of your targeted capabilities and outputs that you're trying to maintain with AI. In our next slide, we'll actually show you an example of this if that's a little nebulous or hard to conceptualize — hopefully that will help firm it up a little bit.

Next up is our AI Model, and I think most people familiar with AI are familiar with the concept of AI models. A great example would be something like Claude Haiku 3.5, which is very much in vogue right now. But there are a number of different types of AI models. Claude Haiku 3.5 of course is a version of an LLM, but there are a number of other AI model types — image generation models, video generation models, and a number of other types. We've got some examples of that as we go forward as well.

It's important to understand that not all models are the same and not all models function the same. It's not just a matter of context window and how much a token costs — there are a number of different capabilities that different models bring to the table, and some of that can be related to the specific model purpose and what it's meant to be doing — a result of how it was trained.

The child records from an AI Model are going to be your Prompts and your Data Sets. For some organizations, managing prompts as assets may be a little bit foreign. It may seem like managing a CMDB CI record as an asset in and of itself. But it's important to understand that prompts, in many cases — especially as case law is coming out about this — are considered intellectual property for many organizations.

Prompts are the set of instructions that tell the model what we want as an output. If you're familiar with any conversational interface with AI, this is the message that you send the AI — that is your prompt. In other cases, if we're talking about systematic AI with integrations or AI that's plugged into other systems, prompts can be generated as an output from deterministic code or a number of other sources.

The reason that we track prompts as assets within this space is we want to know A: how it's performing, B: can we optimize those prompts, and C: what are we using the prompts for, so that we've got a catalog of these things and can actually go back and recreate some of those circumstances and in some cases defend either our compliance with regulations or our intellectual property.

And then finally, Data Sets. Data sets are what's used to train or serve an AI model. In many cases, if you're thinking about your classic Claude scenario, data sets might be something that you are not immediately touching. These could be data sets that the model was trained on prior to you ever coming into contact with the model. So those might not be assets that you actually manage within your organization and within your asset management database. But data sets that you are personally using to train your AI models or to drive certain outputs — those are things that you should consider managing as assets, because of their importance to your IP and the product that these AI systems are outputting.

And then finally, we will also look at managing MCP servers. While these aren't part of the hierarchy that we see here, they are assets that support many of these workflows and many of these systems. If you're not familiar with what an MCP server is, MCP stands for Model Context Protocol. These servers actually help your AI models and systems communicate with one another, both inside and outside of your environment. And as you mature in your AI adoption, they do tend to become very critical to that success.

Here's a quick example of an AI System. Within this example, we're envisioning an AI-enhanced software QA support system. Within this, we've looked at a structure that has several different models supporting the system. We've got Claude Opus 4.8, Claude [MODEL NAME], and Claude Sonnet 4.5, and each one is performing a slightly different activity.

In this case, Sonnet is using the data sets for our previous testing and our development standards, and it's got a prompt to review submitted code and perform unit testing. Opus will then review defects identified in testing and use the data sets for our test results, our previous testing, and training Stack Overflow. And then finally, [MODEL NAME] is reviewing suggested code fixes and considering the architecture.

This is a 30,000-foot view of what a system like this could look like. But the idea here is this is why we are tracking a number of these different components as assets. All of this data is really critical for us. And on each level we may be interested in specific performance components, cost modeling, or any number of things. That's why from a data model perspective, we're interested at every layer of this. Each individual record itself can also pose immense value for your organization.

As you think about building your AI asset strategy, one of the things to keep in mind is understanding key AI asset attributes. We've already touched on a few of these, but obviously it's important to know the model name, the source — which is going to be either a vendor or a provider, however your organization is actually getting that AI model. Your model parameters info — the number of parameters and model training procedures are super important to understand. So understanding whether that's linear regression, deep neural networks, etc. The reason that we want to track that is we want to understand whether one type of training is performing better than the other, especially for in-house models, which are also things that you should be tracking.

We want to know model size. Generally, the ServiceNow AI asset management toolset looks at the size of the model in megabytes. And then our context window — the size of the input sequences that the model can handle. All of these pieces of information are good to know in general, but certainly really critical for us to accurately catalog, categorize, and measure the performance of our AI assets.

Christine Morris

All good stuff. Let's do another poll.

Put your responses in the chat. If you discovered tomorrow that employees are paying for 200 AI subscriptions with company credit cards, your first reaction would be: A: that can't be right. B: let's find out who's paying for what. C: I knew this day was coming. Or D: honestly, it might be more than 200.

We all love our shadow IT folks. Let's see what we got. Jeremy says D. Same with Garrett. Next question is who was paying for it. That's the B's and D's.

Ian Cahall

Pretty spot on.

Christine Morris

I knew this day was coming. And that's the scary thing. That's why what we're talking about here today is so important. It's like the early days of the cloud, when everyone was procuring software outside of normal processes.

Ian Cahall

Next up we're going to talk about managing your AI assets and what that means. Obviously we're going to be focused on the ServiceNow perspective of this. As a result, what we're going to be talking about is ServiceNow's AI Control Tower.

For those of you that aren't familiar, this next segment of the webinar is going to be focused on an introduction to what AI Control Tower is, how it works, etc. As I go through, if you have any inline questions, don't feel shy about asking those in the chat, and either Christine or I will notice and make sure that we get those answered.

Quick introduction. What is ServiceNow AI Control Tower? From here on out you'll probably see that indicated by the AICT abbreviation. AI Control Tower is ServiceNow's all-in-one unified AI adoption management and governance toolset. It uses ServiceNow's workspace interface, which in theory everybody that uses ServiceNow today is familiar with on some level, as most all of the tools have adopted some version of the workspace.

The idea is to make an easy-to-access, easy-to-navigate, and very front-facing toolset that uses the underlying technology that ServiceNow has to connect and discover technologies, bring in information about your AI within your environment, and give you a comprehensive toolset to own and manage and address any deficiencies in your AI strategy. It does that with a number of out-of-the-box AI Service Graph Connectors, which, if you're familiar at all with the ServiceNow platform, Service Graph Connectors are your easy-button integrations. These are the things that allow you to connect very quickly and seamlessly to a number of third-party platforms that give us the types of data we're looking for here. But it also offers an AI Gateway. AI Gateway is a specific toolset that helps manage and oversee MCP servers and a number of agent-to-agent connections to source information on how and where AI is being used within your environment.

From there, we get to layer in all of the good stuff that ServiceNow has been doing for quite some time — on-platform workflows to help manage the work and decision making around your AI technology and your AI assets, as well as dashboarding and data visualizations to help make sure that everybody's fully aware of what's happening, where, and why.

Before we jump into too much of the specific technology of AI Control Tower, I want to talk about the AI Control Tower roles and responsibilities, specifically for asset management. Now these are real-world personas that could be roles assigned to specific individuals within your environment. And of course, there is an analog on the platform for specific roles that we can give users, that give them the capabilities to do their jobs within the platform as well.

The first one is your AI Steward. This is the person that owns and is responsible for your AI strategy. Your organization could have one AI Steward or many. Within the platform, the things we're specifically interested in seeing the AI Steward do is configure and maintain AI Control Tower. Of course, they're going to be responsible for AI discovery — helping us determine where AI is being used and how it's being used, and then hopefully over time drawing that into a conclusion of why, and whether we're happy with that.

They're going to participate in AI asset management activities — creating assets, approving assets that have been submitted for creation, and making sure that it is gaining value for the organization. And then finally, they're going to be responsible for collaboration with your cross-functional teams to help support things like adoption and AI success.

On the other side, we've got our AI Asset Owner. If you've been in our previous MasterClass sessions for asset management, this is a role that is very common across all of our other asset categories. Very often we'll have a dedicated hardware asset manager or a dedicated software asset manager, or potentially multiple. This is the same idea. This person is responsible for making sure that we've got those AI asset records in ServiceNow — they're up to date, well kept, and maintained. They're going to manage and execute the asset lifecycle that we just talked about a couple of minutes ago. And they're going to make sure that they're using AI Control Tower to drive AI value and adoption, so that AI assets are being utilized appropriately — or sunset, if not — so that we're not wasting spend and resources on tools that we're not invested in making a success. Both of these roles have an actual on-platform ServiceNow role that you can give users to make use of these specific capabilities within AI Control Tower.

When we think about AI Control Tower, one of the key things it has the ability to do is provide visibility into the AI asset inventory. There's an actual tab within the core screen. And what you'll see is a view very similar to what you see here. You're going to get visibility into your AI asset inventory by specific category.

Just like we talked about earlier with the data model, you're going to see how many AI Systems, how many AI Models, how many Prompts, and how many Data Sets you have today in total. And of course, like always with any workspace view, you will be able to drill down into those. You can see those by type. When we think about type, we do have Agentic AI, Generative AI, and Classic AI. Given how the last couple of years have gone, those categories and types will only continue to branch out and evolve.

And then we've got our AI Systems by stage. Just like we talked about in the previous slide around roles and responsibilities, we see that we've got 100 AI Systems up for Steward review. That means they've been submitted, but somebody actually needs to go through that assessment process and approve them so that they can proceed through development — that's our Build and Test cycle that we talked about. Things that are approved for development will begin work so that we can actually bring them in. When they complete building and testing, they can go into deployment. And then ultimately we can see and manage how many AI Systems have been deployed. And then we can see that broken down by department and other things. The idea here is that we're getting a lot of visibility into the current lifecycle state and quality of our asset inventory. Through this view, we can drill down and get much more information on these items.

We've got a question: "Will we have the ability to make changes to this AI Control Tower workspace UI to tailor the interface for different branches of the business, like a dashboard? Or is this workspace fixed in its configuration?"

Great question. There is a default setting for this workspace. For those of you that aren't familiar with this, ServiceNow has been gradually rolling out configurable workspaces and the ability to make modifications to these. What you see here is a relatively static view, but there are capabilities to create configured workspaces that would give you a more tailored experience if that's something you're looking for. While what we're looking at here is relatively static — outside of the ability to apply certain filtering — you do have the ability to create a configured workspace that would give you the control you're looking for.

Next up we've got the AI Control Tower Value Management tab. This is where we are tracking the overall value of our AI. There are a few key things that we're trying to track when we qualify what value means for AI. There are a lot of different best practices that are coming to the surface as this AI space gets more and more fleshed out. Some of the key KPIs we're looking at up front are going to be things like productivity hours — an estimation of how much time AI has saved your organization.

The number of users is also very critical. We can apply components here like the number of specific actions that AI is taking within your environment, the number of unique users, and systems by value — a calculation of which systems are providing the most productivity across the most users, etc.

When we think about this value, it is an assessment somewhat removed from the financial piece of the equation. But that's not something we can completely ignore. Within each of our AI System records we do have the ability to actually track things like token spend. We can track what those tokens actually mean in terms of product efficiency, etc.

And one of the things that we also recommend — and this is something that AI Control Tower does have the ability to do in addition to AI asset management — is monitor individual token performance or prompt performance. That's where we get to see how frequently prompts have to be resubmitted to get to what is considered an approved or complete output. These are all qualities that ServiceNow is trying to surface for you based on the data and the integrations it has available. We'll talk a little bit more about the integrations piece of this here in just a moment.

But before we get to that, I do want to talk about how that AI asset lifecycle process works within AI Control Tower. Previously we talked about what that AI asset lifecycle looks like. ServiceNow does have within AI Control Tower a tailored experience to help drive that lifecycle forward. You can see it here in the UI on the right-hand side of the screen.

Once we've onboarded an AI asset — where we're creating this ServiceNow NowLM 1.0 record and defining some of the core elements of that record, saying this is the source, this comes from ServiceNow, here's the model name, here's the context window, here's the size of the model, all of those sorts of things — we're then going to initiate this asset lifecycle.

Once this model has been saved and submitted, it's going to start the assessment process. This Assess phase out of the box generally has some default approvals that can be assigned to your AI Steward or other members of your organization. But this can actually be configured to support specific requirements in addition. If we need to spin out an assessment for your TPRM process, and you're using TPRM on the platform, you do have the ability to do that. If you need to have somebody answer a questionnaire that's not connected to the TPRM process, that can be a part of this lifecycle stage. In order to complete this step, somebody will have to go through whatever the assessment criteria is, complete the task associated with it, submit it, and close it complete in order to move on to the next stage.

In doing so, you get a number of guardrails implemented out of the box that bring you some of the governance you're looking for in this space. Number one, you're going to get a record of that approval. Number two, if there are any questionnaires or those kinds of things happening, you're going to have the record of that assessment completed and associated with the AI model we're talking about. That way you can always access it in the future should you be audited or should there be compliance questions, etc.

From there we go into the Build and Test phase, and this is again going to give you out-of-the-box tasking for development of the AI asset, model, and system, but also testing of that development. While this isn't a full replacement for a complete software development lifecycle toolset, a number of users are currently doing that on ServiceNow anyway. If you're doing your software development lifecycle outside of ServiceNow — maybe you're in GitHub or Jira or Azure DevOps — no matter where you are, you can associate that process with what's happening here within the asset lifecycle, so that we're clear on how those actions are occurring and that everybody's on the same page with where we're at in the asset lifecycle.

Once we've completed that testing, we can go into Deployment, and here we can actually schedule and execute those deployments — up to and including automation of deployment from ServiceNow if we want to. That could be assigning subscriptions — maybe we're talking about a Claude Pro subscription, or maybe we're talking about deploying an agent on individual users' devices. We can do those things with a built deployment plan that gives us visibility into that. And then of course, once we've completed that deployment, we'll have visibility into which of our AI assets are deployed, and we'll be able to actually manage and maintain them until we're ready to offboard them in the future, should that ever come.

Also important to call out — offboarding can be initiated at any step. If we don't pass the assessment, we can go ahead and offboard the AI asset. If we don't pass testing, we can offboard it at that point as well. We do have a lot of flexibility here. But the idea is to have some structure, much like we would with any of our hardware or software assets.

Next up, let's talk about integrations. This is where the rubber really meets the road. It's one thing to say that we can add manual asset records and those sorts of things. That's not really how we look to do things from a ServiceNow perspective or in general from a best practice standpoint. We want to automate as much as we can, and obviously even more so when we're talking about AI.

As we talked about before, AI Gateway does allow for integration via MCP servers. It does that through one of a few different processes. AI Control Tower has its own native UI for using and creating MCP servers and creating connectivity with those if they exist outside of the ServiceNow space. You can also do so through ServiceNow's AI Agent Studio, which is a different toolset that we don't have time to get into today. But if you're familiar, it does give you the ability to create and manage MCP servers. And then there's the MCP Catalog, which is a toolset that actually shows you a number of available vendor MCP servers that are out there that you can integrate with ServiceNow to bring in external AI data.

The idea here — while we're talking about ServiceNow and the on-platform things, our example previously was NowLM — this is really meant to be comprehensive for all of your AI, especially those that do not exist within the ServiceNow ecosystem. We're thinking about things like AWS, Salesforce, you name it. We can bring in all of that AI. We'll talk about some of the Service Graph Connectors that are available for that, which is our other path to bringing in AI data.

For those familiar with those standard integrations, generally speaking, with AI Service Graph Connectors, we are looking for something as simple as an OAuth token and permission to integrate. And those third-party platforms can give us a ton of good data about what our AI assets are on that third-party platform, what they're doing, when they're doing them, and give us the ability to actually execute decisions around those — hey, how do we want to trim these back, do we want to set a new budget, etc.

Here's a quick showcase of what those available AI Service Graph Connectors are today. You'll notice on the whole, this is pretty much a who's who of the major Western AI providers today — AWS, Databricks, GCP, Hugging Face, IBM, LangGraph, Microsoft, Moveworks — although Moveworks will ultimately be migrated onto the Now Platform — OpenAI, OCI, Salesforce, and Snowflake. More are coming. We know that just based on what we've heard from ServiceNow so far. If the third-party provider you use is not currently on this list, it doesn't mean that we can't get the data — it just means that either we need to go through an MCP server like we talked about previously, or we may need to do a custom API integration, which is not the end of the world either.

As you're thinking about which AI systems you use or systems that you could be using in the future, it's good to have this mapped in your mind as far as what we're going to have to do to get that data over to ServiceNow. And of course, all of this is enhanced as we think about the evolution of ServiceNow over recent years — bringing in RaptorDB Pro, Agent Fabric, and Data Fabric — all of which help make sure that this data moves much faster and is much more performant on the platform.

I want to touch on some of the non-ITAM AI Control Tower capabilities. In addition to the things that we've talked about today, AI Control Tower also brings to the table a number of other AI management capabilities. With all of those assets brought in and you managing that asset lifecycle and the performance of those things, we also have the ability to secure our AI — we can manage our AI agent identities, our access controls, make sure that we're establishing our privacy baselines and reporting around those. We can manage our AI performance — in addition to the financial and adoption components, we can look at things like reliability and efficiency. We can establish workflow connectivity with other areas of the platform like ITSM, etc. We can track organizational adoption and actually gain AI-driven insights into how to improve our AI.

Meaning we can use ServiceNow's Now Assist capabilities or its autonomous capabilities to look at our AI estate and make decisions around where we can do better and where we can improve things. And then in addition to that, of course, we can do support for our AI — we can open cases to track enhancements and user feedback, open issues to track impacted users and resolutions if there are issues with a deployed AI technology, and create requests for AI assets to drive change or adoption if needed.

Christine Morris

All right, our last poll. Do you know where to start with AI Control Tower? A: yes, in fact we've already started. B: yes, we just haven't started yet. C: I have a good sense of where to start. D: no clue.

Some folks have a good sense of where to start. Has anyone started yet? Lots of C's.

Ian Cahall

For those of you that are looking for where to go — and I see we did get a D in there — we do have some things that will likely help with that.

When we think about where to start with AI Control Tower and as an extension of AI asset management in general, it's important to understand the why first. All the things that we just talked about — it all sounds great. We think about asset management as something we should do, it's a best practice. But deeper than that, we've got actual specific reasons why AI asset management is really important. We talked about shadow AI previously, and based on your responses, there are probably people out there who are subscribing to their own AI and doing things on their own. It is a little bit of the Wild West, just like cloud was 8 or 9 years ago. We're kind of re-experiencing that now with AI.

There's a lot of risk around that — whether that's legal risk or cost risk. We want to make sure that we're addressing that. In the EU, the EU AI Act was passed. There are deadlines and audit findings that are going to be required around that. Generally, the EU is several years ahead in terms of regulations in comparison with North America. But we can basically look at that to say this is probably coming our way sooner than later.

And even if there aren't external regulatory factors, we know that for organizations using AI today, we want to know what it's doing for us — what is it costing, how is it performing. We need to be able to demonstrate and show an ROI. Having that data is critical. And I think above all else, we've seen how rapidly this space has changed just in the last two and a half years since ChatGPT really hit prominence. As these things are coming into fruition and as we're rolling out this technology, the speed with which they evolve and the speed with which they act means that we have to be really controlled around how we work in this space.

For those of you that aren't familiar, ServiceNow has included AI Control Tower in their new SKUs that just rolled out a couple of months ago at Knowledge. But for a number of legacy customers, it is also possible that you can access AI Control Tower at no cost. So it's a really good time to look into this and consider whether this is something that's on your roadmap for this year or next year. And obviously the criticality of that is going to be related to how far along you are in your current AI adoption.

We do have a solution for this. A number of organizations are doing this in different capacities today. We've introduced the Ondaro Lighthouse for AI Control Tower. Lighthouse is an expert-led 90-minute discovery session where we work with you to understand what your AI environment looks like today in a rough sense — what kind of telemetry you have, what kind of risks and ROI you have on the table — and bring you back a tailored activation plan for AI Control Tower. The idea is that this will rapidly get you to that endpoint of having a good grasp on what your asset inventory and estate looks like and how you can do the most with it.

It's a straightforward process — we take you through this four-step plan of knowing what you've got, having a good understanding of it, applying governance, and then ultimately driving that optimization through this Lighthouse offering. As we work with our customers and try to figure out what AI means for them and how they can make the most of it, Lighthouse has become our guiding path to help customers get there.

Next steps — I'll turn it over to you, Christine, to walk us through this.

Christine Morris

Let's ask the audience — any questions or comments? Has this been helpful? Give us a thumbs up if you liked it. We're constantly looking to evolve these sessions and give you what's meaningful to your business today. Lots of thumbs up — love it. If you are interested in doing the Lighthouse Discovery session, definitely reach out to us. It's quick and relatively painless, but it will really start to get you on a great foundation for moving forward.

In addition to that, if your organization is looking to start using things like Now Assist or AI, we do have an AI Spark offering. One of the things we've seen most frequently is folks who have been on the platform for years who have customized heavily. Ten years ago, ServiceNow said you can do anything you want, and a lot of us did. So it's critically important as you start this AI journey on the platform to have that good foundation.

One of the things that we do is come in and clean things up so that you can really start seeing value. What we've seen from clients who just dive in and turn it on is that some of those customizations are preventing that value. It's always a good idea to have somebody come in and figure out how to phase this in and start to see that value. And of course, we've also got our Ondaro Reserve offering — all of our consultants are AI-trained. And then just a quick reminder, if you are interested in looking at some of our other topics, we posted this early on in the chat, but you can go out and take a look at our other MasterClasses.

We appreciate everyone joining today. It was an exciting topic and kudos to Ian for doing a fabulous job. The final message is: let's get ahead of the governance of it. Let's start measuring it, let's start monitoring it. We've seen in some of those news stories where token costs are astronomical and it's sneaking up on people. Really understanding the value that you can get and monitoring that is going to be critical for folks to be successful going forward.

Thank you all for joining. Stay tuned for our next installment. If there are topics you can think of, you're welcome to put those in the chat for some of our future sessions. And of course, you can always reply to any of our emails with feedback as well.

Ian Cahall

Thanks very much.

Christine Morris

Everybody have a great day.