Unlocking AI potential: What you need to know about ServiceNow Yokohama release

Sean Dawson: Hello, and welcome once again to the Cask Distillery Podcast, where we unlock the full potential of ServiceNow with expert insights and practical strategies only here on the Cask Distillery Podcast. And speaking of expert insights, I have with me today Nate Weldon, who is the senior VP of technology. And we're going to be talking about-something new that we're doing as a series is talking about Yokohama new release and the AI capability highlights that are coming along with that and get into that.

But Nate, thank you so much for taking time out of your busy schedule to do this with us. Really appreciate it.

Nate Weldon: Oh, you're quite welcome, Sean. I'm very excited to be here today.

Sean Dawson: I wanted you to give your background, because you've been here forever. And I thought it would be good for you to go over your background, because you'll do it fast.

Nate Weldon: Well, forever certainly makes me feel a little old, but I have been here for a little while. So, sure. So, I've been with Cask for almost 11 years now. Prior to Cask, I was with a life sciences company in Southern California, and that's where I got introduced to ServiceNow. So, I've been a customer of ServiceNow for six years and a partner for almost 11. So I've got about 17 years of experience with the platform. 

Before I got involved in ServiceNow, I was on a CIO career track where I was starting in help desk infrastructure, app dev, and ERP. So my background is primarily in IT operations and leadership.

Sean Dawson: Well thank you. I wanted to talk a little bit about Yokohama in general is the release. It has a significant meaning. And ServiceNow's Yokohama release is really named after, really, the city, which is a dynamic port in Japan. And it symbolizes innovation and global connectivity, which, as you'll see as we talk about this more and more today, for those listening or watching-however you're consuming this-you'll see that that has links into what this is doing with AI and all that. And I love that piece of it. 

And Yokohama, being a major port city in Japan, is also known as the gateway to the world, representing innovation and global connectivity. So it's a really, really cool tie-in with what we're doing today with AI capabilities and highlights coming in Yokohama. So Nate, could you talk to us about some of the most exciting AI capabilities that are going to be possible now with the Yokohama release?

Nate Weldon: Sure. So obviously, with all the activity out there, everybody knows what's coming with agent AI or agentic AI. Everybody's doing it these days. But what ServiceNow has to offer, I think, is going to be a little bit next level than what else we're seeing out there in the marketplace today. Because ServiceNow is also going to be releasing Workflow Data Fabric. The power you're going to get is not just being able to use data or information from your ServiceNow platform to help your agents make good, sound business decisions, but you're also going to be able to pull in data from your various other silos, data lakes, and virtually any place else you might have data residing in your enterprise environment. And in some of those instances, you won't even have to bring that data into ServiceNow with what they're calling the zero copy data connector capability. So, the power is just in how much data you can use to allow your agents to make good, sound business decisions on your behalf. 

So the additional power you're going to get on top of all of that-when you're looking at agents and how you're going to design your agents and run them-is the controller capability that ServiceNow is also going to offer as part of this release. So, think about a crawl, walk, run approach that Cask very regularly advises for our customers: You don't want the agent to make bad decisions on your behalf and then potentially upset some of your customers if you're using this for CSM or some of the other external non-employee functions. 

But even with your employee functions, you want to take what's probably a supervised approach first and make sure that the AI or the agent and all the tools underneath it are returning good, sound recommendations and good data because you don't want to be the victim of what we call a hallucination and have some bad information be put out there, either internally or externally. Hallucinations aren't good. And we want to make sure we're avoiding all of those things, especially when we're dealing with things that can impact our business reputation, our security, our customers' opinion of our products. 

So we're going to be advising a structured approach and, again, a crawl, walk, run approach that is more supervised at first. And once you start to feel comfortable that you've educated your agent and you've educated your AI, and ALMs are there, then you can move into more of an autonomous AI agent structure. But really, the three capabilities together are, I think, going to revolutionize what you can do with AI, agentic AI, generative AI-all of those things are going to start to come together and coalesce around this ServiceNow product suite.

I'm a firm believer that there's 99 percent of business problems that can be solved in technology. And ServiceNow is absolutely the platform to be that solver of those business problems. So I'm very excited about what's coming.That was what was released yesterday, and the future iterations of these products, because I see there is tremendous power and capability. And ServiceNow is the enterprise AI provider that I think most organizations are going to be willing to accept and trust with their data. It's about trust. It's about making sure bad things don't happen because we didn't pay attention to it well enough. And ServiceNow is just that platform to make sure we don't make those missteps in the future.

Sean Dawson: And you mentioned something called a hallucination. So I wanted to hit on that. For those that aren't familiar with AI and the terms behind that, can you describe that a little further just what that is?

Nate Weldon: Sure. So, those of us that are using some of the ChatGPT or Google's Gemini, sometimes the responses you get, you know those are false. You know those are incorrect. So sometimes when the AI is generating its response, it can, for all intents and purposes, make stuff up. So we want to avoid that. We don't want an agent making stuff up on our behalf that could impact our business.

Sean Dawson: And the other thing that you brought up is the agentic AI, and I wanted to see if you could talk to our clients-or the people listening, rather-on what clients need to understand about agentic AI and the capability behind it.

Nate Weldon: Sure. So I think first and foremost is you want to go slow. This is new. Anytime something is new, we have to go slow before-you got to slow down to go fast. So, take a measured approach first. Understand what your business goals and what you want to accomplish, and plan these things out before you even touch ServiceNow. Know what you want the end result to be. In addition, for some of the other capabilities, you might need to rely on some data that should already be in your ServiceNow environment. 

But we do see plenty of situations where incident resolution is one or two words-"I fixed it"-or the description of the incident is "Emails broke." And although a human probably knows what to do with that, the agent or the generative AI controller needs more data. In addition, if your knowledge base is not populated with well-structured or at least good content in your knowledge base, you're going to be limited in what you can use AI for just because ServiceNow still needs something to learn from to be able to generate a case summary or a knowledge base article.

So, you want to look at what your historical data is. And, sure, you could use an AI agent to even beef all that data up, but you probably want to go back and validate and make sure everything looks good.

Sean Dawson: And I know there's an agent and a skill. So what's the difference between an agent and a skill?

Nate Weldon: So, skills are basically single-use case things that you can do with ServiceNow-things like, I built one that can generate acceptance criteria. Right? So you take your "I want this result so that I can get this business outcome," and a single skill will generate that for you.

Agents take a little bit of a step further, and you can link skills together. You can use multiple LLMs as part of a single agent process or agent workflow and really start to get to a lot more capability and a lot more autonomous decision-making once you've ensured that the AI is giving you the right data back the first time. So again, back to that supervised versus autonomous approach.

But agents will leverage skills as well as other platform capabilities like Workflow Data Fabric. But an agent could leverage a flow, a sub flow, it could leverage a decision table. There's numerous different other things that ServiceNow could leverage as part of that.

Sean Dawson: And you mentioned Workflow Data Fabric, which was the next thing I wanted to touch on. What is Workflow Data Fabric, and why does it matter?

Nate Weldon: Sure. So, Workflow Data Fabric is a suite of tools. Some of those tools have been out in the product for a long time as part of the Integration Hub product. But Workflow Data Fabric is a is a suite of capabilities that will also include zero copy data connector and some other capabilities for bringing data from your various distributed IT systems into ServiceNow for use with an agent, a skill, but also you could use it as part of a flow or a sub flow if you wanted to.

Sean Dawson: Okay. And you had mentioned earlier-we were talking about data and pulling in data from different things, especially with the Workflow Data Fabric. But how does quality of data play into this whole thing?

Nate Weldon: Quality is paramount. I mean, there's the age-old "garbage in, garbage out" concept. When we used to talk to our customers about bringing in historical data from legacy systems, the traditional method is export it to Excel and then import it into ServiceNow, and everything should be good to go, right? But one of the worst things you can do is bring bad data into ServiceNow. You want to call that out first. 

So data quality is still paramount. You have to have good data-and generative AI can make up for some poor data components. It's not going to solve all of that. You do still need good data. Especially if you're not using publicly accessible LLMs. If you are only using your data-like if you're using an internal LLM-then absolutely the quality of that data is paramount because there's nothing else to learn from. There's no ChatGPT to reach out to as a sounding board to say, "Hey, does this thing sound right?"

Sean Dawson: So what do you see as the barriers for clients to take advantage of these new capabilities in Yokohama?

Nate Weldon: Well, there's not a lot of barriers to agentic AI. If you're on a Pro Plus or enterprise queue, you are going to get this capability as soon as you upgrade to Yokohama. So I guess that's the first barrier is you have to upgrade to Yokohama. So if you've got upgrade challenges and you've got a lot of technical debt, and you're basically adhering to the N-2 Support model, you're not going to get any of this until you upgrade to Yokohama. So that's barrier number one. 

Barrier number two I think is good-quality historical data. So in the machine learning capabilities that predated all of this agentic-I keep wanting to say generic, but that's not the correct word; generative is the correct word. So I apologize for that. But there were AI capabilities that predated both of those, and machine learning is one of them. You still had to have 30,000 records of data to teach the machine how to categorize your incidents, how to automatically assign the incident to the right spot. So that hasn't changed. You still need good foundational data for the AI to look at to understand what good looks like.

So again, if you've got one-word descriptions and one-word incident resolutions and vague problems and change records that don't really give you much context, I don't think that's going to be very valuable to your AI evolution.

Sean Dawson: Yep. So from an action standpoint-again, for the listeners and people watching-what's the best way for people to get started with this?

Nate Weldon: I mean, besides calling us? Try to figure out your business problems that you want to solve first. Don't just turn it on. And you could look at some of the out-of-the-box agents, and some of those may be very useful for you. And if they are, great. Test them, obviously. And you can start to release them into production and use them more comprehensively in your organization.

But if there isn't an agent built already, think about that business problem you're trying to solve first. Think about where the different data that you need to make a decision on that business process is going to live and reside and can you get to it to bring it into ServiceNow. Chances are the answer is yes.

But if you've got security and compliance requirements that say you're not allowed to integrate with it, then that's not a data point that you can use for your AI decision-making process. So understanding your environment, knowing your application infrastructure and where your data lives and resides-but as well as the compliance and the access restrictions behind those, you need to understand all of that before you start building an agent.

Sean Dawson: Yep. Well, great. Well, thanks again, Nate, for the time. That's all the time we have for this today. And really appreciate it.

Nate Weldon: All right. You're quite welcome. I'm looking forward to chatting again. 

Sean Dawson: Yeah. So for those of you watching, again, like, subscribe, and let us know what you want to hear about. And we'll be doing more of these new releases as things come out to give you a CMA's expert view on this. Nate's one of them. And we welcome you to provide us feedback, and we'll talk to you later. Bye bye.