Awesome. Welcome, everyone. Thanks for joining us to talk about generative AI today: fact, fiction, and the future of work. We have so many wonderful attendees here. Let us know where you're calling in from, dialing in from. Put a note in the chat from where you're at in the world today. We're excited to have so many wonderful attendees with us. My name is Sabina Varani. I am a product marketing manager here at ServiceNow, focused on our customer workflows, and I have the pleasure of being joined by Neil Kostecki, who is the Director of Outbound Product Management, also with customer workflows. Welcome, Neil.
Hey, glad to be here. Thanks so much. Super excited to talk about this topic. It's a hot one. Generative AI is on everybody's mind. I'm Director of outbound product management for CSM. And I've actually been here for just over a year. So I'm overseeing the CSM portfolio. But before that, I spent a lot of time in service. I led a product team at one of our build partners. So AI, search, all of that goodness. And before that even, led a global KCS practice, so Knowledge-Centered Service, if you're not familiar. But contact center, service, AI, search, all of this is near and dear to my heart. So happy to be here.
Excited to have you here with all of your expertise to help all of our attendees from around the world. We have folks coming in from Germany, from Los Angeles, Mexico, Connecticut. I think we are going to have such a great conversation today. Some of us are freezing cold. Some of us are nice and sunny and warm, but let's get into what we have going on for us today, so we're all here to learn about what is real, what is fact, what is fiction as it relates to generative AI. We'll tackle five myths about generative AI as it relates to customer service. We'll look at what's the future look like for us as we look at where we are today and then where are we going. Some key takeaways to leave our audience with, and then we'll get into Q&A.
Please feel free to drop your questions throughout the conversation today in the Q&A box. We'll get to as many as we can. And for those that we can't get to today, we'll be sure to follow up with you afterwards. So let's get into it. Our first myth that we hear is generative AI is a security and compliance risk. So how are you ensuring the implementation of generative AI is secure? Is it compliant? What guidance are we giving customers as it relates to AI? Neil, what are your thoughts?
Yeah, this is absolutely the number one thing that we tend to start conversations with customers and partners with is really, before we start leveraging AI of any sort, we've got to think about security. We've got to think about the way that we're leveraging it. And it is important to think about how you're doing this. As you can see here, you don't want to pick kind of that patchwork approach where you're leveraging point solutions, trying to stitch things together. And you can end up with scenarios where people are copying and pasting information from one system into another and then copying it back.
And so then you get this information that's in this transit that really isn't managed and secured with a platform that gives you the ability to bring in and connect to any type of model, any type of experience that you want to embed that into as far as workflow. It really gives you the ability to have oversight and to have an understanding of how you're using it and making sure that it's super secure.
So what you're saying is if you're taking this copy and paste approach of different point solutions, that patchwork approach, that's when you introduce more security and risk issues into your kind of day-to-day operations. But if you're taking a platform approach where everything is integrated, you're getting that layer of compliance across your organization. That's how you kind of get that win-win-win of getting more from your technology, cutting your cost, and bottom line, working smarter across the board.
Exactly. I mean, the key to remember here is generative AI is generating content. And so one of the things you always see when you're interacting with this across the platform is verify for accuracy. So we're making sure that you're aware that you're engaging with generative AI. We're providing the guidance to make sure that you're verifying. It should always be human in the loop. So that's an area where we're always trying to ensure that we're providing accurate information. But trust that information and verify what you're reading and before you leverage it.
Right. Giving that human touch to it, right? It's not generative AI gone rogue. We have to have that human oversight element to it.
Totally.
Amazing. So another thing that we're also hearing when it comes to generative AI is that AI in general is very costly to do an implementation as it relates to customer service. And it also can be quite complex. Sometimes it can take half a year to a year. You have to get stakeholder buy-in. How are you approaching this as kind of you're looking at implementing generative AI for customer service?
It's funny you asked. Literally yesterday, I did a webinar on how to activate your analysis for CSM and walk through the steps of how to actually deploy an out-of-the-box capability. So case summarization, for example. Super easy to do for business users, for admins that can really follow a guided step-by-step dropdowns, radio buttons. We're not talking development here. So that's number one is being able to just instantly deploy these out-of-the-box capabilities into the workflow. Second is our job is to make sure that we are continuously innovating, that we are providing you value. This is SaaS. You're subscribing to continuous innovation. So that's an area where you're always going to get new functionality. Every single quarter, we're releasing new capabilities to the ServiceNow Store. And so especially being able to go out of the box means you're going to get that value quickly.
It's a single platform, again. It's a single platform with AI across the entire ecosystem. It's not even just when we're talking about customer service, obviously, that's a no-brainer. It's in all of the areas of your customer service. It's also across all those other use cases across the ServiceNow platform. I think really it's definitely sure you can go complex, but you don't have to.
Especially as it sounds like having a low-code and no-code approach to deploying with those radio buttons that you're talking about, having it ready out of the box helps mitigate some of that complexity. What I also heard from you is that you're investing not just in what you're doing today, but building it on a platform that has continuous innovation so that you're not just subscribing to a one-time implementation. Then six months, a year, two years down the line, you have to redo your implementation, have it connect to your legacy systems. It seems like that would be not just complex and costly, but a short-term fix. It seems like having this out-of-the-box solution and a platform that grows with you as your organization grows could be helpful.
Absolutely.
And what I'm also hearing about kind of here, Neil, is that we've heard it from you. But do you have kind of some metrics or a case study that you can reference from a customer that's gone through an implementation?
Yeah. So you'll see here, this is our customer Kainos who have implemented CSM Pro, CSM Pro Plus. And they're using generative AI today. And you can see some of the metrics that they're getting here: a 99% customer satisfaction with a low-code, no-code implementation, 17% increase in operational efficiency. And the big one here, 71% reduction in resolution time. That's a significant time savings. It's a significant dollar savings. It's also a great experience for customers, for your agents. One of the things that is really key here is that not only have they been able to have significant improvements to their experience and for their agents, but it's also allowed them to lower the amount of headcount growth that they need with the amount of customers that they're bringing on. So they're really able to do more without having to bring on as much as they used to.
Actually, we've got an interesting quote here from Peter Pacher, who's director of application management and strategic growth at Kainos. He said, "It frees up time for everyone to focus on future development and take us further down the path we're on." So I think the key there is it's helping them achieve their goals. Their goals aren't to deploy and necessarily manage a GenAI solution. Their goal is to drive their business and have great outcomes for their customers. So it's enabling them to do that.
So what I'm hearing is with Kainos, it looks like not only was their implementation able to save them costs in the long run, but it allows them to do what they do best, which is focusing on their line of work. And we at ServiceNow can continue to innovate on our platform, deliver these low-code, no-code solutions to achieve a user-friendly interface and scale with that organization.
Absolutely.
Amazing. All right, let's move on to our next myth, which is when we think of AI, it is a machine of some sort. And so these responses can be impersonal and robotic. So if you are trying to adopt it, especially in customer service when you are facing frontline customers, how do you ensure that when you're using AI, you don't have this impersonal and robotic experience?
Yeah, I think we're not talking about the old chatbot framework that everyone's kind of familiar with, the experience that you might have had before. We're not talking about better chatbots. We're talking about leveraging LLMs, leveraging generative AI. You can really get to a more conversational tone. You can really provide really rich content that is very convincingly human. It needs that human approval. But I mean, it's really hard to tell. It's why there's a lot of challenges in, let's say, the education space where you've got students who are trying to write their essays and their content with that. So it's a whole different level. So we are talking very personal, very rich conversational type responses. We're also looking at scenarios where we can leverage something like the knowledge graph.
So the knowledge graph is this way to contextualize all of your information about you, your location, the services you have, the product you have, the data model that lives around you in kind of your ecosystem as a customer, as a customer of one of our customers. And we can leverage that information as part of the conversation. So reduce the amount of effort that's required to actually insert that information or the amount of questions you're going to get asked through a chatbot, through the AI agent or virtual agent type experience. Just fill that information in. We already have it. We already understand it. So that's, I think, another really key aspect.
So it's not just a better chatbot. It's not just another sequence of scripted or canned responses that we've probably seen in the early days of AI, of virtual agents or chatbots. We're moving towards agents who can understand and personalize before a customer has to input, "OK, this is my order history," or, "These are challenges that I've had before." It seems like with generative AI and AI agents, there's this additional component that makes them not just a chatbot, but more conversational, personalized. And you mentioned a little bit about knowledge graph. Can you talk a little bit about kind of what that looks like or what we're thinking about from the evolution of AI, kind of what you're seeing today?
Yeah, I mean, Knowledge Graph fits into kind of our whole data strategy. And you might have heard about Workflow Data Fabric. And this is how you really connect all of the information that you have in your organization and bring it together and leverage it to draw insights, to drive agentic workflows. So it's absolutely a key aspect of how we use that underlying data. I think when you talk about AI, you cannot not talk about data. The unstructured and structured data is really the basis on which you're able to draw those insights, generate content, get the context that is needed to have a great personalized conversation in a virtual agent, for example.
Right. And that kind of lives on that Workflow Data Fabric layer, which is built into the platform, which is kind of what we were talking about earlier as it relates to why it's important to have this platform approach, not have a patchwork approach when you're looking at your AI and Generative AI strategy as an organization.
Absolutely, yes.
Got it. All right. Thanks, everyone, for continuing to put questions in the chat. We'll continue to answer them as we go, and we have time at the end as well. We'll move on to our fourth myth that we're hearing a lot from folks just like you, and you just talked about data Neil. You talked about how it's important that we have kind of organized and quality data in order to have AI be useful, but there's this myth around AI models needing vast amounts of customer data to be useful, and we're wondering, what's the reality of that? Do we need vast models? How much data is needed in order to get started, and key here is to actually be useful for the organization.
Yeah. I mean, as far as the amount of data, we've got large language models, but we've also got small language models. You really need to think about, I think, before anything, one of the most important things is thinking about what problem are you trying to solve. That's number one. And then you start thinking about the models and the tasks or the workflows that you might need. But we think about industry-specific scenarios where we need to understand a specific type of data. So these are areas where having a flexible approach and leveraging different size models for different approaches is super important. Our models, we have our Now LLM service. It's built into the platform. We're able to leverage that for a number of different use cases. We also have the ability to connect to third-party models.
So bringing them in through what we call the Generative AI Controller. So you can connect to any of the models that you see out in the space. You might have your own pre-trained model that you can connect to. So I think there's different models for different approaches. I think it's really important to think about that first. And the security of being able to connect those models into the platform and manage them and really have the visibility and governance over them is also super important.
So looking at it as having pre-trained models, having that specificity allows the data that's on these models to be useful for these customer use cases. Is that what I'm hearing?
Exactly.
OK. And I think another thing that we're hearing too, seeing in the chat, is back to this notion around, OK, if you have an AI model that can be more conversational, how do you ensure that there is not this perception that the AI is all-knowing and omniscient and is a human indeed? I think you mentioned something earlier about how we have to take it with the lens of verification as well. Is there something that we kind of preclude or put as a warning when we're using AI?
Yeah, absolutely. Everywhere that you see in the platform, you're going to see that indication to verify the accuracy. And I mean, I think the thing is everybody is used to, I'm assuming there's a lot of folks on the call that are used to engaging with some form of generative AI in their day-to-day. It might even just be in your personal life. And you have to know that, number one, you're engaging with something that's pulling content and generating an answer and that you need to verify the information. And you'll see citations. So when we provide an answer through a virtual agent, you're going to see the answer, but you're going to see also these numbers that will indicate the source document that you can go back and read and just make sure that you understand the full context.
So yeah, it's super relevant and important to verify. But yeah, I think it's really about when we're talking about models and the size of data, again, it's really dependent on the use case. We have use cases for code. We have use cases for generating knowledge articles. We have use cases for summarizing cases. So generating a playbook, it's an entirely different thing. You're creating a workflow for agents to go through. It has nothing to do necessarily with text. It's generating a structured process for an agent to follow. So different models, different amounts of data, different use cases. It's all relative.
As I understand it too, when we're looking at models itself, we have our Workflow Data Fabric. I also understand right now that there's a lot changing as it comes to what models will be used, what will be allowed. Maybe we can talk about this as we get to the future, but is there a strategy that ServiceNow is taking around what models we are using and what we connect into?
Yes. Yeah, I mean, it's really, like I said, we have our Now LLM service, which is models built on our infrastructure that we own. But we also have the ability to connect to, for example, Microsoft Azure, OpenAI. You can connect to Amazon Bedrock. You can connect to your own pre-trained models that you've got and bring all of those through that Generative AI Controller. We'll continue to expand that. I mean, the platform is designed to be flexible and to be able to provide that flexibility, but also the security and oversight. So being able to manage and understand what models you're using for which skills. So you can actually go in and see, OK, I've created a custom skill through Now Assist Skill Kit. It's addressing a specific use case in my business. Which model is underneath that?
That's important to be able to keep, to manage and to understand and have visibility on. So the more that you bring in, the more that you're leveraging, the more ability to see and share and have folks being able to have ownership and accountability over that.
When you're connecting to these additional LLMs, my first thought is, OK, are we introducing security or are we introducing risks back in if we're taking additional LLMs? How do you ensure kind of back to our original myth here of this can be a risk? How do you ensure compliance? How do you ensure that there is scalability too as you're moving in with this?
Right, so we have our own as a platform. We have our own AI Governance, how we deploy, manage, and we have white papers and information that you can find on the website around how we do that. Anytime you engage with a software vendor, you're going through some terms and agreements around what the platform is providing to you and how they govern their AI. So when you start introducing third-party models, you're also engaging with that vendor as well in one way or another. If you're, let's say, Microsoft Azure or OpenAI, you are engaging with that vendor and you're making that agreement around how you're going to leverage it and what their practices are around governance. Then you're also integrating it through the ServiceNow Generative AI Controller, which is the approved and secure way to connect to that model.
So you're really going back to the number one, which is the first thing people always worry about is risk. Well, the way to reduce that risk is to connect in these ways rather than trying to kind of custom integrate, make API calls, doing it in a way that is not built into the platform and leveraging the tools that are provided. I think that's really an area where just using the tools to connect to those third-party models is going to not only do it in a secure way, give you the oversight, it's also going to ensure that maybe you have a use case in customer service and you had to connect to a third-party model. Now, other parts of the organization can also leverage that in a secure and scalable way.
Because it's kind of come in through that controller layer on the platform that we're talking about.
Exactly.
OK, great. Thanks, Neil. All right, we have a lot of questions in our chat. We'll keep them coming. We're going to get to our last myth that we have that we've heard not just from people who are in the AI space, but people who are new to it, people who are seasoned veterans as well. And that is that generative AI will replace human agents. This is a fear. This is, in some cases, a reality. From your perspective, Neil, what are we looking at when we're looking at generative AI, the human workforce? We know we have a global talent shortage. Is generative AI going to help in that by replacing humans? How is that going to work?
This is an interesting one. And it's the first thing that comes up when we start talking about AI agents. It's like, OK, well, it's just going to replace everybody and everybody's going to be out of jobs. That's not the case. I mean, I think everybody is engaging. I mentioned before, most people are using some form of generative AI, Copilot, something in their personal life. And you haven't been replaced. What it's done for you is helped you get the skeleton of an idea to start with a first draft to help kind of remove that writer's block or that moment of like, I'm stuck. I need some information to kind of bounce off of. And so I think it's enabling humans to be able to do better work, to get to better results faster. It's what's the word I'm looking for? It is augmenting our abilities.
We talked about generative AI and skills and how that's providing agents with more productivity. AI agents are here to work with humans also to help augment some of those tasks that are areas that are more mundane. I mean, being able to provide more conversational experiences, being able to do those summarizations in chat, calls, all of these things help to provide more productivity and more focus on what's really important, which is servicing the customers.
How exactly is generative AI kind of working as it relates to knowledge base or knowledge articles, especially as it relates to kind of both the customer experience and the agent experience?
Well, I mean, the ability to create a knowledge article with a single click is kind of just mind-blowing. Having managed a KCS practice, and really the hardest part often in getting a knowledge management program off the ground is just getting people to write and being confident in writing that first draft. It's honestly really difficult. People struggle with having perfect grammar and having everything well-written and having the whole thing sorted out. And it takes time. And you're in a contact center. That next call is waiting for you. And your after-call time is going up.
Being able to just take a case that you just managed that has the information written inside of it, the context of what you did, being able to just click a button, generate a draft that's very well-written, that's brought all the information in, and then make a few changes, it's incredible. It goes full circle because that knowledge article is going to help the next, let's say, 100 agents that are trying to resolve similar issues. Then that article is going to become published to the knowledge base, which is also going to become available to self-service. Customers will not only find it, but they'll also get summarized generated answers through their search experience. I mean, that's, I think, the biggest thing that generative AI has really done is been able to provide really well-written, clear, context-driven content. It's pretty powerful.
So what you're saying is some of these kind of tedious tasks that agents might have had to do before, they're trying to balance the incoming calls, the additional cases coming in their queue with also doing the wrap-up notes. Let's say they had an issue that they know could be helpful for the future, but to take the time to go through a mundane task of summarizing everything, doing it in a clear and cohesive way, and then saving it, that whole kind of skeleton of the draft gets there faster through generative AI. It still requires that human oversight, that check to say, OK, yes, this works. But now it's helping not just future agents, but it's future-proofing your business because you're able to take an experience of one agent, help multiple agents.
And then that also goes back into the self-service portal so that, let's say, a customer runs into a similar issue in the future. Well, because we have generative AI, the customer doesn't have to go through the knowledge base and search, OK, which one of these are relevant? What I'm hearing is that generative AI will comb through the knowledge base and go ahead and surface up the most relevant knowledge article, maybe in a summary format, and even cite, kind of like you said earlier, which sources it uses so that, let's say, the customer wants to go in deeper and read those specific articles. It's already served up to them. Am I getting that right?
Exactly. Yeah. And I mean, recently, we're talking about knowledge content. But in our Q1 release, we also released SharePoint and Confluence connectors. And this is an area where we're continuing to develop. Probably folks on the call might have heard about our acquisition of Raytion, which gives us the ability to connect to more content sources. So this is providing you more of that context that can really provide the right answer for your customers.
Amazing. Well, thanks, Neil. I think we've had all of our five myths debunked today. And as we're kind of looking towards the future, we're hearing a lot about AI agents. We saw it in the press recently. And I'm curious to get your thoughts on the difference between, as we've talked about, those virtual agents and what are these ServiceNow AI agents? What can they do? What can't they do? What's the reality here?
Definitely. This is, I think, more so than just generative AI. AI agents is a super hot topic right now, and the way to think about this is it's really, again, a human that is working with a digital AI agent that's able to help them get work done faster and more effectively, so in this case, an AI agent that can help a service agent that has a number of low-priority cases that might need follow-up, that need more context from the customer. It might need to provide some knowledge content or an answer that might be helpful for the customer, and so really, it can do that autonomously, and then when it runs into a scenario where it needs some approval, there's a you have the ability, when you define these AI agents, to be either supervised or autonomous.
So if you have a scenario where it's supervised, it can come back and say, hey, I need approval before I can take this next step. So it's a whole nother level. And I think it's going to really unlock a lot of productivity for our customers. Like we mentioned, we're not just talking about the better chatbot here.
So what we're looking at is these agents, in a sense, can act autonomously to help get work done, but there's still that human component. And as we're looking at it from a broader landscape view, can you break it down for us a little bit more in terms of what are the roles within agentic AI?
Yeah, so it's important to understand that we're talking about a team of agents, so the difference here is not just individual skills like summarization or generation. What we're talking about when we think about AI agents is a team of agents which are led by an AI Agent Orchestrator who has the ability to say who as though it's a person, so the AI Agent Orchestrator is able to reason, to come up with a plan, to actually use the context and information that's been provided to then say, OK, I have a plan. I know how to proceed with this plan. I have a team of agents. Each of those AI agents has a specific purpose and skill, a set of tools, and one of them might be able to look into order information. Another one might be able to send email communications.
Another one might be able to verify the context of a case. So each one of them has their unique purpose. And that orchestrator knows each of their purposes and knows which one to call on in order to achieve its plan. And it's going to go through that and get to resolution or reach out to the human agent to say, hey, at this step, I need approval before I continue. And it's going to learn as it works through those cases, let's say, what was a good outcome, what was a bad outcome, and get better over time. So it's really think of it as a use case, let's say, resolving a case. And then you have those individual team members that are going to help to get to that case resolution.
These team members are specialized in their one or two skills. So that's why you need a team of AI agents to complete the use case with that human approval that the AI agent orchestrator oversees kind of who's doing what. Then it will have to go sometimes to a human to say, OK, does this make sense? Did this go through correctly? Here's a summary of those steps that generative AI has provided.
Exactly. Yeah.
So this is a role, it sounds like, the AI Agent Orchestrator and the human element component. This is kind of new to the ecosystem for. This didn't exist before when we had just virtual agents. Am I getting that correctly?
Yep. Yeah. If you look at our new and we don't have it in the slide, but if you think of our new kind of AI layer, it's really AI agents and the AI Agent Studio. So how do you create them and kind of manage them? There is the AI Agent Control Tower, which really lets you have oversight over everything as far as agents on the platform, agents that are third party, and then you have visibility around understanding your agentic workflows and understanding how they're progressing and basically having oversight. It's important to be able to not just implement an agent, kind of set it free and hope that it's doing the work, but actually being able to manage and see everything in that ecosystem. We're going to have, I mean, hundreds of agents we're talking about here.
For a single use case, you're going to have four or five. In some cases, you might have up to 10 or something. So very quickly, you're going to see that there's going to be a lot of AI agents that are able to also be reused. So reusable AI agents that are kind of at the platform level that you might want to use in your own use cases. So we've definitely upped our game on the ServiceNow AI agents part of our platform. Really excited. The release is coming up really soon. It feels like tomorrow, but it's just in March. So we're really excited for this.
As I understand that, these AI agents are slightly different in the fact that they're available 24/7. So it means that you can have support, especially for customer service where it's imperative to get those case resolution times down and to get customers the answers they need quickly. To have this available around the clock, it seems like that's a huge win in terms of customer service, but also on the value that AI agents can bring so that as a company scales, you're able to continue to service your customers without necessarily having to increase headcount at the same level. You'll still do it, but in a way where now the new hires might be focused on learning about AI Agent Studio and ensuring that there's that oversight there to make sure that the team of AI agents is working for the customer, for the humans.
Yeah. I mean, if I think about that human-agent interaction and going back to the will we be replaced, there's a lot of concerns. I remember when I was kind of in the knowledge management space, a lot of organizations that were concerned about putting their knowledge in a self-service forum. Like, oh no, if we put the self-service content out there, the customers will be able to self-serve. We won't need our contact center agents. We'll just close down the contact center. All the solutions will be there. That is absolutely not the case.
I think there's always going to be this evolution of, OK, we implement some AI agents that are able to handle some of these specific tasks that today are repetitive, mundane, and that could be automated, that just might need a bit of reasoning, which is why you're leveraging an AI agent, which is going to give your human agents the ability to manage highly complex, highly valuable interactions where you also need to not only have the complexity and the knowledge to be able to solve and troubleshoot and address these issues, but you also want to build a relationship with your customers. I don't think there's any organization that is going to want to put their customer service completely on autopilot and not have any humans available to help build that loyalty, build that experience with their customer base.
I think it's going to be an evolution, and there's going to be a shift in how we do support in a contact center. But this is going to be a slow process, I think. We're going to start to experiment with specific use cases. Customers are going to have their own. We have out-of-the-box use cases. And it's going to be progression. It's going to be an evolution. And I think I'm in the business. I'm working with the product. I'm not concerned about this replacement. I'm excited about the augmentation. I'm excited about what will be possible to be able to actually have a team of agents managing those low-priority cases, providing approvals.
You'll have folks that are then able to maybe there's a new role that comes out of it that's actually someone who their job is to for this team of agents that's managing that, they're actually the official owner of that. Like you said, new skill set. They have to understand how to use the AI Agent Studio. They need to understand how to engage with that platform, so I think it's exciting, and we don't know how things will evolve, but I think I'm optimistic.
We're ready to go. We're already on it. Well, we've talked about a lot today, Neil. I wanted to just kind of bring us back to what are the key takeaways that we want to leave our audience with. And so one of the kind of highlights and themes that I've heard is that AI is here. AI, generative AI, and just get started. Start where you are and just start. Get on that train because it is an evolution, and you don't want to get left behind. Get started with getting familiar with what are the capabilities, what's available to you right out of the box on your platform, and just get going. I know that sometimes we want to go with 99% or 100% certainty, but it seems like this is one of those technologies that we learn as we go.
The technology itself adapts with our learning as well.
Definitely. I mean, do it, I would say, like all of the things we talked about, the security and the platform and everything that's there to help you start quickly. You're not starting necessarily scrappy with a point custom solution that's connected just through APIs or something, but adopt what's available, adopt it in an out-of-the-box way, and start small. Don't be worried about your entire strategy, but really start somewhere, learn, expand. The sooner you start getting familiar with it, the better you're going to be able to leverage it.
Absolutely. The other thing that I heard a lot was around security, compliance, and risk. And I know that we've had a lot of questions in the chat as well around how do you ensure that it is secure? And it takes us back to this theme of having a platform approach over that patchwork approach of copy and pasting from various different data sources and then not knowing if the customer data is secure. So having that platform approach, what you mentioned earlier, having that Workflow Data Fabric layer as well to have the right data ensures security. It ensures efficiency as well because, as we know, this takes a lot of power, and it takes a lot of cost in there. So you want to see where can you cut costs and also gain those efficiencies.
Definitely.
Yeah. And then lastly, what we heard was that AI agents need humans. It seems like they're not going to be replacing humans but complementing the work that humans are already doing and freeing up those humans to do the more complex strategic work that you can't necessarily have an automated agent complete. There still needs to be that relationship development. There still needs to be sometimes these complex tasks that even if you put 500 AI agents on it, it still won't get done in an efficient way that a human can do it.
Yeah. I think it's important to know that AI agents is not a single agent that has the ability to do every single task and do everything that a human agent would be able to do. There are specific use cases that you define. And within each case, you'll have an orchestrator that knows how to orchestrate for that use case. And you're going to define many use cases potentially for service, many use cases for, let's say, not in my space, but in sales and order management and in areas in field service. The idea is that you will define and identify these cases where you have a very specific scenario that you have the workflows and the tools available to automate that and automate it potentially with some approval and guidance along the way from a human.
You're not going to stand up a service support AI agent that is just going to handle every case, period, that your customer service organization is handling today. So it's important to know that, again, there's that evolution. You're going to start to build those out. You're going to start to find the areas where there's high value. And also, just going back to talking about accuracy. As you build out your AI agents, and this is something we'll start to have guidance around how to build an AI agent, what are the things you need to think about, is certain areas of that process might be completely deterministic. I know for a fact that it's almost rule-based, so that AI agent can go do that stuff. Don't approval. But there's going to be scenarios where there's some uncertainty, some ambiguousness.
That is an area where you're going to need to have that supervision approval. In some areas, it might mean that that's not a use case for an AI agent. All things that need to be thought about as you start to embrace and enable this technology.
Right. Because at the core of this, we're using AI for people. We have to remember that as we go forth. And we have that as our charter in a way to say that, yes, we can build all of these great tools and capabilities, but at the core of it all is that we're putting AI to work for people. And so we need to make sure that we have that human element and that human thread and perspective as we're going through this. So like you mentioned, there might be some use cases that are, let's say, you're approving an expense report, and it falls within these very strict rules of, OK, it's this level and this amount of a report, then it can approve it.
It can go through the workflow, and then it can give the summary to the human agent to say, hey, this is what I did. This is how it's approved, but then there are those more complex cases that require a bit more of a strategic brain, and AI agents are specialized. Even if you have a team, even if you have that orchestrator there, it still isn't a great use case or a great use of computing power in order to have it be an AI agent use case, and we're going back to having AI be there as a complement, but not a replacement because we still need human agents. Great. All right, well, we are going to move on to our Q&A, and we have a lot of questions in the chat. I know that some of them have been answered as we've been going.
And I see one in here that says, recent developments in AI have shown that you can create an extra-large model on a very low budget and small amounts of data, and that can still be useful or valuable. So why go with ServiceNow if we're seeing that in other parts of the world?
Right. I mean, as far as AI developments in general, this is something that is literally happening right now. Every single day, every minute, it's innovation. It's AI. There are teams and I don't know how many millions of people that are all working on this. And that's our job also is to do research and to know the latest and greatest and to invest in a strategy and build that out for our customers. So you're always going to see some new innovation coming across the newsfeed around something that has changed in AI. And I think what's important is to know that we're tasked to do that for you. That is our promise to you is to innovate for our customers to provide you the best value for what you're trying to solve for.
And as an example, we're able to connect to those third-party models that you might want to leverage. And that's an innovation in how we connect to and secure and let you leverage those. But there was a really interesting talk I just want to mention that I went to at one of our World Forum events. It was our VP of Research, Nicolas Chapados. And one really interesting thing he said was, when you look at everything that's happened with LLMs, one of the things that people in this field didn't really expect is that just building a bigger, better next word predictor model was going to change the game. And so that's someone that, oh, that's all they do. They're in this space. Many people are in this space. And they didn't necessarily see that coming.
So there's going to be constant innovation. The pace is difficult to keep up with as a customer. As a customer of ours who's trying to whatever, they're a retail location or they're in finance, that's not your main role is to try and keep an eye and tabs on every AI innovation. That's definitely ours. So yeah, that's why I would say just don't get distracted by the fact that there's always going to be innovation. Embrace it and learn about it and be aware of it. I think it's important to do that.
Makes sense. Thanks, Neil. I want to take us to another question that we hear is, how can organizations balance centralized governance and control while encouraging innovation, ensuring rapid adoption while maintaining security, compliance, scalability, and interoperability?
Yeah. So in Q1, we released what we call AI governance. It actually gives you the ability to have oversight over all the models, all the skills, everything that you've implemented, and also provide the ability to have approvals so that you can actually design the approvals and have them assigned throughout your organization. So you have the oversight. You have the visibility. You have the ability to have approval and workflows that are assigned to making sure that before a new model is leveraged, that it's gone through the proper approvals. So again, that's really us building into the platform the ability for you to not just have to think about that as a process and a strategy, how you're going to manage your GenAI, but actually giving you the tooling and oversight directly in the platform.
So that's what I would say is a great feature, definitely worth checking out. If you've installed the Q1 store update, you're going to have access to that.
Got it. Thank you. And then one other question. I know that we're getting close to time here, as it seems like there's a lot coming out in the world of AI. Where do you go as someone who lives and breathes this every day? Where do you go to stay up to date on the latest trends and news for generative AI? And with that, I'm going to give you a two-part question, is, if someone were to try to get started, like we talked about, we want them to get started, but if they were to get started today, where should they first start? So two parts here. Where do you go to stay up to date? And then where do you go to get started? Or what's the first thing that you would do?
Right. So I would say as far as trying to stay up to date, I mean, I'm constantly looking at what's going on in the space. And my newsfeed is full of it. I've got my Apple News feed is bringing me AI news. I'm going through colleagues that are sharing great stories and innovations through my social network. I'm basically kind of crawling all over. And I'm also using generative AI to do the same. I'm leveraging generative AI to actually learn about generative AI, which is also a great way to do it as well. But we have also research that's going on in our organization. So there's ServiceNow AI research that's published. There's a lot of information out there. And so it's almost impossible to not be learning about new things.
It's a great thing to do to constantly keep yourself informed and explore some of these new technologies. I think how do you get started? What's the thing that you should really do? I think, well, first of all, you're here. So you're interested in the topic. You're learning about it. You're getting some ideas around things that you should be thinking of. We have plenty of content about Now Assist. You're going to see AI agent content out there and a lot more coming. So we have lots of great Live on ServiceNow sessions. We have documentation. But I would say, yeah, just look for the content we have that's like get started. Make a plan of action of the problems that you're needing to solve in your organization. You shouldn't really start, I would say, doing anything until you know, hey, we have a problem here.
Being able to address this would be really helpful. It does help you define how you leverage AI. I think Now Assist Skill Kit is a way to build a skill for any use case with any data with any context you want. Think about the problem you're trying to solve and then work your way through installing, getting started with out of the box, and then explore some new skills that are going to get value for your company.
Amazing. So it seems like there are resources out there to help you get started. I also know that AI and agentic AI is moving every day. There's something new coming out, and so what I'm also hearing in the feedback from the Q&A is that people are wanting more of this, how to get started, what support is available, what education is available so people can self-serve, and there are these live on ServiceNow webinars that are going to continue to come out to support in that education aspect, but this is great feedback for us to hear so that we know what else we can continue to provide, and so please keep telling us whether it's on this forum, on our communities. Feel free to reach out afterwards, but we want to hear on what is it that you want? What content do you want to see?
Thank you so much, Neil, for all of your insights. It's been a pleasure to hear from someone who works in this space day in and day out. I know that we have many webinars to offer everyone here, not just the one that you see today, which will be available on demand, but we have more on-demand webinars available at the link provided here. I just want to remind everyone to please fill out the survey that you'll be getting to let us know how we are doing, the feedback. We take this feedback seriously. We read each and every comment. We try to ensure that we're taking your feedback as we plan for the next webinar, the next engagement, so that we're giving you exactly what you need. With that, thank you to everyone who joined us today from around the world.
We've loved spending this time with you. We didn't get to all of your questions, and so we'll be sure to follow up with you offline and answer those questions one-on-one, but for everyone who joined in, thank you so much for your time, your attention, and your engagement, and have a great rest of your morning, afternoon, or evening from wherever you are.