(upbeat music) ♪ Yeah, yeah ♪ ♪ I'm on a roll and I'm ready to go ♪ ♪ Turn up the heat and make those embers glow ♪ ♪ Yeah ♪ ♪ I'm on my way and I don't skip a beat ♪ ♪ I can't help that I've got happy feet ♪ ♪ I'm lighter than air and I'm feeling so good ♪ ♪ I do what I want and it's understood ♪ ♪ There's no stopping me when I'm doing my thing ♪ ♪ I'm doing my thing ♪ ♪ I'ma do me ♪ ♪ You know what I mean ♪ ♪ I'm on fire ♪ ♪ I'm feeling glad, there's a spark in my soul ♪ ♪ I light it up, burning brighter than gold ♪ ♪ Break it down now ♪ ♪ I'm getting happy, feel my heart go zing ♪ ♪ I'm clapping my hands and I'm pulling these strings ♪ ♪ No stopping me when I'm doing my thing ♪ ♪ I'ma do me, I'ma do me ♪ ♪ You know what I mean, you know what I mean ♪ Please welcome ServiceNow's Senior Vice President, Platform and AI, Jon Sigler. (cheerful music) - Good afternoon, Knowledge. My name is Jon Sigler. And when I requested this meeting, they told me we would never get enough people in this room. So thank you very much for coming. We are gonna drill down on the platform and AI in our platform. But before we get started, I wanted to follow up on something Bill was saying yesterday. These are unprecedented times in technology. The internet, that was pretty big. The iPhone, that was pretty big. generative AI is huge. IDC predicts that in 2027, the AI market will be $521 billion. I think that it's a very large number, but I think it's too small. I think it will be more than that. 80%, over 80% of corporations are spending more on AI this year. Our early adopters, customers that have purchased generative AI and the ServiceNow platform, are already seeing a 3.5X return on their investment, which is incredible, but that number is still going to go up. Generative AI is not gonna replace jobs. But somebody working at a company and that company does not adopt generative AI and the Now platform for business transformation will be left behind. So we've said this a few times this week. AI is only as powerful as the platform it's built into. And that's extremely important. Calling out to an AI service is not gonna get it done. You have to build the technology into the platform. generative AI is really good at intent, it understands what we're asking it. It's really good at synthesizing and summarizing. And it's really, really good at language generation. But that's where it stops. It turns out that ServiceNow has the world's best automation platform, and what that means is we can take action and we can get things done. So when you combine AI plus our platform, you get real business transformation and you get it quickly. So we have been leveraging AI in our platform for seven or eight years. And we have customers that are using those features and they're getting real value out of it. We have things like doc intel and task intel and AI search and predictive intelligence and all these things, and they're great. But as I said in the beginning, it's nothing compared to what we're gonna see with generative AI. So today we're gonna go deep, we're gonna dive deep and look under the hood and we're gonna show you what's in market today. And not only show you, but we're gonna tell you how it works. Then we're gonna go and say, here's what the rest of 2024 brings and what you're gonna see in our products and our platform. And then we're gonna see the art of the possible and we're gonna look out into the future and we're gonna see what might be possible in the future. So before I bring out my first speaker, I just wanna say, Heath Ramsey is behind almost all of the live demos that you're seeing in these keynotes, Heath Ramsey and his team. And that's a lot of stress and that's a lot of work. So I wanna introduce Heath Ramsey, the hardest working man at Knowledge. Heath? - Thank you. That's a heck of an introduction, and one that I was not expecting. Very happy to be here today. Very happy to be talking to you about generative AI. As Jon mentioned, I am Heath Ramsey. I'm the Vice President of Outbound Product Management here at ServiceNow. I have a pretty amazing job. I get to talk to people about generative AI. I get to help people understand what it is. And hopefully, get everybody to implement it as quickly as we possibly can. I've been with ServiceNow approximately 11 years at this point. And as Jon mentioned, I have never seen anything as transformative as what we see right now. And I gave my first talk about generative AI back in March of 2023 before we even had a product. But what we figured out very early on is that it's not about generative AI, it's about the workflow. And the workflows are so important. And generative AI needs to be built into the platform in such a way that those workflows can take advantage of it in a very intelligent kind of fashion. And that's exactly what we did by making generative AI and Now Assist, which is the experiences that are powered with generative AI, available throughout the platform at that platform level. Any solution that is built on top of the platform is Now Assist-aware and generative AI-aware. We might provide that to you out of the box natively, ITSM, CSM, HR. You might build that with creative workflows. But at the core of it, you're able to tap into powerful technology to do all of those different things that you needed to do to remain competitive. So what do you need? You need data. If you're a longtime customer, even if you're a short-time customer, you have a lot of that. You need models, AI built for work, NVIDIA, Watsonx, Hugging Face, the models that we're providing that are out of the box and available to you. But the most important thing is the actionability, making sure that people can actually work and do meaningful things with generative AI in the context of ServiceNow. See that Watsonx right there in the middle? We have a great partnership with IBM. And they are also looking at data and models and everything else and working with ServiceNow. And I have a little message from them for you. Please roll the video. - Clients are going through an unusual time where they've gone from, they need to run their business and do AI. We've been through eras of data science, machine learning, deep learning. But the difference is it's getting easier to work with, easier to integrate into your workflows. That's why as we think about what it takes to be successful with AI, it's more about what are the data sets that you can bring? How do you do training? And a big focus of ours, how do you make the economics work for a client to be successful with AI? So where's the architecture go? Well, as I like to say, there is no AI without IA, meaning information architecture. Data is going to decide the competitive advantage and the winners in AI. That's why we're really excited to be working with ServiceNow to see the power of ServiceNow powering your workflows along with Watsonx delivering unique insights and capability. Thank you. (audience applauding) - Said very, very well, the marriage of data, marriage of workflows. But let's go a little bit deeper in terms of how we're gonna be able to do that. So this morning you saw in CJ's keynote, my apologies, one slide back please. You saw in CJ's keynote the knowledge graph. It may sound a little bit esoteric, but it's actually something that's really important. You've got all of the data in your instances. You've got relationships between a user, between services, between integrations and systems and all those different things. Those relationships change over time. I travel, I'm here in Las Vegas, I go to a lot of different ServiceNow offices. When I go to a different office, I should get a different set of services potentially. Instead of me having to notify someone that I'm there, we just know that. We know that relationship, we know that it exists, we know that it's changed. That's what the knowledge graph does, and it's going to drive incredibly personalized experiences. Now what about the models? This really is about the best model for the job. What do we mean by this? As I mentioned before, we're going to give you plenty of models, plenty of capabilities, plenty of skills that are going to power your solutions. But you know what? You're also going to need to bring other models. You may have something that you're training, you may have something that's competitive to you. And so our DNA is built on top of interoperability. Our DNA is in integrations. Our DNA is meeting people where they work. And with our generative AI strategy, there's no difference. So you're gonna be able to bring these models within your organization and power experiences natively inside of the ServiceNow platform. But it doesn't mean anything if people can't do work. Every customer that I talk to talks about making sure that there's some kind of quantifiable value. Okay, that's great, but how do you quantify it? Well, in ServiceNow we quantify it in work. This is what we do. But what it means for us is that we have to pivot our roadmap to personas, thinking about how people are doing work in the platform. We're figuratively putting ourselves into the developer, the expert, the end user roles and trying to understand what does it mean from an experience perspective? And so we think about outcomes because we think about what that particular person is trying to do. Automation backlog, how do we help people automate faster? I know that you all have massive amounts of backlog that you're trying to reduce, ways that you're trying to digitally transform. And more importantly, how do we put those capabilities and skills into the subject matter experts? This morning you saw image to playbook generation. Those are SMEs on a whiteboard that are able to build processes. They know the business. They can help you transform better. On the expert side, agents, operators, analysts. How do we remove the tedium? How many people really wanna create another KB article? Raise your hand. I don't see any in the audience right now. But these are all things that help people be more productive, get up to speed quicker, find the answers that they need. And of course the self-service of customers and employees, critically important. I should be connected with the information that I need when I need it. And not just information, I should also be connected with workflows. I should be connected with all of that entire catalog of things that you've built to be able to drive outcomes. And you know what, that's exactly what we do and it's what we're gonna show you within the next section. But before we get there, I'd like to share with you someone who has started this journey around Now Assist and generative AI with ServiceNow. And I would love to bring to the stage the Chief Information Officer of TRIMEDX, Brad Jobe. So good to see you Brad. Thank you for coming. So Brad and I, we probably had our first conversations in the fall of last year. But before we get to that, tell us a little bit about TRIMEDX. - Yeah, so TRIMEDX is a clinical asset management company. We really focus on clinical engineering, so taking care of equipment inside a hospital. We have a clinical asset informatics business which helps hospitals understand what assets they have and how they're being utilized. And we have a cyber business that clearly is trying to minimize cyber risk in the hospitals. - Yep. So when you started looking at generative AI in the news, what was your first reaction to making that something part of TRIMEDX? - I think it's just like everybody here. You know, the excitement, the opportunity, the problems that it's gonna solve. Plus the extra push from the executive team, from the board. Why aren't we doing this, we need to go faster. However, you know, we're a healthcare company at the end of the day. And the actions that our technicians in the field take have the ability to impact patient outcomes. And so we're still charging ahead, but doing it in a cautious way, and trying to make sure that whatever we do, it's not a patient outcome that we're impacting. - Yeah, and as a prospective patient of yours one day, I appreciate that conservative approach. But even though you have a low risk portfolio in terms of some of those outcomes, you still have the opportunity to explore. What are some of the opportunities that you did find for generative AI in Now Assist? - Yeah, so I think I can lump it into two categories. We have, number one is reduce our cost to serve, right? Increase productivity of our technicians. If you look at the population of 3,000 technicians out there, they complete about 2.5 million work orders every year, open about 300,000 purchase orders. If we can reduce five minutes out of that, that's a real ROI. If we can do it by taking away administrative work, now it's also an engagement tool because we both know they hate the administrative side of it. So now it's an ROI and an engagement tool. The other piece of it is, you know, we have what I'll call an aging population. I can say that 'cause I'm probably the same age. I got gray in my beard too, so. - But in the next 10 years we're gonna have a real problem where these technicians start retiring. And so you take that individual that has 35 or 40 years of experience, how do we consolidate all of that knowledge, all of the secret sauce tricks he or she has and put it in that 22 year old's brain in some way? So that's a big use case that we're looking forward to. - Well, and I love that because that's like, organizational workforce transformation at the same time, and you're thinking about it very, very proactively. But of course you're probably gonna start a little bit smaller, and I think that was with Now Assist for Creator. Tell us a little bit about what you're doing with that. - Yeah, absolutely. So in our world, right, we wanna push forward and really tackle this. So we started with my development team. I've got 22 ServiceNow developers. And as soon as we turned it on, we had five that immediately started using the text to code. And you know, they're the early adopters of the technology. Just turn it on, no training, no nothing, away they go. What we've seen is about a 22% increase in their productivity by doing that. So that starts through word of mouth and other things, right, increasing adoption along the way. So a couple months later we're about 50% of the developers are using it. And since I think half of them are here, you know, we're gonna have to bump that productivity number up. But I mean, that's really where we're at right now. - No, and I like that. It's find the people who can be promoters, the folks who have the most affinity to it and they really can lead the way, and it's kind of like through osmosis, which is good. So if you were to just quickly summarize some of your key takeaways and learnings from, you know, your starting of the journey and where you're going, that would be great to share with the audience here. - Yeah, I mean, you need to find your early adopters, right? You need the promoters who are gonna take it, run, take the good, the bad and figure out how to work through it and you know, make it a usable solution, as well as kind of push the adoption wider. You need to have some measures in place to look at, I mean, I heard you talking earlier, is there an ROI? None of us have an unlimited budget to go, we're going all gen AI and there's no return. So we have to be able to measure that and really understand. And that really is around feedback mechanisms, right? Whether it's financial or subjective of how the tools work. - Yes, absolutely. Well, I really appreciate you taking the time to share your experiences with me and the audience, and definitely looking forward to our continued collaboration. Brad, thank you so much. - Thanks. - Okay, so we've talked a bit about the platform, we've talked about some experiences. Thank you again, Brad. Now it's time to get into the cool stuff. We're gonna get into the demos here. And what we're going to do is we're gonna show you the different things that we're planning on releasing in 2024. And so we'll take you through a journey through these different personas. And we are going to start by bringing to stage a good friend of mine, the Senior Vice President of Product of the Now Platform here at ServiceNow, Kush Panchbhai. - Thank you Heath, and thank you Brad for being such an amazing customer. What you heard from Brad is what we are hearing from all our customers. Customers want to reduce IT backlog. They want to create and configure workflows. And they want to create custom apps. And we want to help them with that. We wanna supercharge the productivity of their developers and admins with the power of Now Assist. Last year we released code generation, flow generation, next best action with flow. And we have seen massive productivity. 26% higher throughput per developer. 50% faster coding with Code Assist. Now these are amazing numbers, but we are not gonna stop here. We are gonna add more innovation so that your speed of innovation increases and you get maximum value from the Now platform. To see all the innovations in action, let's welcome Brittany who's gonna help me with the demos. And let's switch to the demos. So here's the secret. When I'm not doing my day job at ServiceNow, I have a side hustle going on. I work on Kush Travel Company and I help people travel across the world to their favorite destinations. But my company is having a hard time right now. If you look closely, my CSAT is going down. And it's going down by 8%. And the reason the CSAT is going down is because the P1 cases of lost items is increasing. Now last year I was at Amsterdam and I lost my bag, and it was not a good experience. And I have a very customer-obsessed culture in Kush Travel Company, and I want to make a really good experience even in a bad situation for the customers. So let's build something amazing. Let's dig deep into that data. Let's first ask Now Assist what kind of luggage or bags people are losing. Now behind the scenes, Now Assist will go and churn through all that data and will bring insights to me. Now I could have done it too, but it would've taken me hours to come to the same conclusion. But Now Assist found this amazing data for me and is giving me this data. Now when I see data like this, my first inclination goes to, I need to create a service catalog for it. Now what is a service catalog? Service catalog is a very intuitive UI through which a customer can request a service from my company. So let's build that. Let's build a service catalog. Let's tell Now Assist to help us with that. And again, within minutes, Now Assist is gonna churn through the entire system and build us this amazing service catalog. Now, building this service catalog again is not hard for me, but figuring out the right questions, the right order would've taken me days. But this is helping me do my job within minutes. Let's review the service catalog and let's look at the questions. You see, it figures out that the customer needs to tell us which trip the bag was lost, the description of the bag, where do they want us to return the bag to? So all the right questions. I like it. Let's go to review and submit and let's hit submit. Within minutes we are progressing and we are making amazing innovations. Now the next thing, which is also super cool, is Now Assist is not only doing things I'm asking it to do, but it's also telling me the next best thing I should do. So it understands that when anyone creates a service catalog, you add automation into it. And that's why it's telling me, would you like to create a new playbook? So let's hit yes. Now what's a playbook? It's a step-by-step orchestration of a very, very complex workflow, an end-to-end process. And it could be across various different departments. And this is exactly what we will build today. So let's go and give it a name. And let's give it a prompt. The prompt is whenever someone loses a bag, let's create a case first. Let's give the task to the grounds crew so that they understand and find that bag. And then at the same time give task to our customer service department so that they're keeping that customer in the loop and eventually returning that bag to them. That's the final outcome we all are looking for. So let's hit generate. Now behind the scenes the large language models are churning and building an end-to-end playbook. Let's open it and review it. Now here you see it created that customer request, it created that ground task. And it's not only creating empty skeleton, it's pre-filling it with existing platform components, whether it's list, forms, your sub actions, your sub flows, everything which understands from the instance, it's putting it in this workflow. Let's scroll and let's review it. It has a decision tree. It understands what to do when the baggage is found so that it tells the customer service the steps for that. And when the baggage is not found, what to do. Let's go to the board view. As a developer, I always wanna see the order of operation. And the board view gives me that view where I can see the different swim lanes and I can see that order. And this all looks good, I love it. Let's go back. And like I say, we have a very customer obsessed culture at Kush Travel Company. So the one step which Now Assist is asking me to do, let's work on that. Let's send a hospitality package. And Now Assist is already giving me recommendations on what I could do. But we all are in Vegas, and I wanna do something Vegas-style. So let's give it a prompt. to deliver a package. And if the customer is already on the return flight, deliver it to their home address. And if not, deliver it to the address which is in the catalog. Let's hit submit. And again, it's creating a subflow for me. We can go and review that subflow. It's an end-to-end flow which is created, ready to go. I can review it, it understands how to do different decisions and come to that outcome. But wait a minute, it's not Vegas-style. It's an amazing workflow. There is nothing Vegas style. So let's do something there. Let's open Now Assist, and let's tell Now Assist to do a drone delivery to the customer wherever they are. That's it, hit submit. Now what's happening behind the scene is Now Assist is going and figuring out that I don't have any integration to a drone delivery service. So it goes and finds that integration, it finds the open API specification for that integration, and uses that open API specification to create an integration hub spoke, all of it done within minutes. I can review the actions and I can click request drone delivery and also cancel a delivery in case something bad happens. Create and install. Just like that, I was able to integrate to a third party system and get a drone delivery service integrated into my flow. Now you all must be thinking that Now Assist is amazing at creating net new objects. How about editing an existing code on the system? So let's go to the response handler of this delivery service and let's review the code. As you can quickly see, the code says kilograms instead of pounds. And I wanna change that because my customers are in United States. So let's select that entire code block and let's say edit with Now Assist. And let's give it the prompt to convert kilograms to pounds. Now as soon as Now Assist is working on it, it understands the formula to convert, it makes the code changes, it even formats it for me so I understand what lines got deleted, what lines got added, and I love it. Let's accept it, let's publish it, and let's go back to the playbook. And within minutes we built an end-to-end workflow. We identified the problem, we solved the problem with an automation, we integrated with a third party system, and we even edited that code, all of it within 10 minutes. Now you all must be thinking that like, when is this all gonna light up so that my developers and admins get this amazing productivity? So all of this will be launching in 2024. Every quarter we are adding new innovations so that you can adopt them and build more innovations on the platform. Now let's see this playbook in action. So for that, I wanna invite Heath back to see how this end-to-end playbook works in a customer service scenario. - Well, I think that was pretty amazing. Do you think that was amazing? Yeah, that is revolutionary. So let's talk about what this now means for the people that are gonna interact with that. Well, with Now Assist in general, it is about productivity. It is about making sure, as I mentioned before, that people are able to do their work faster and that time is saved. And time is money. Look at that 20 plus minutes saved per major incident. Major incidents represent service interruptions. That is literally 20 to 30 minutes that service is back online faster. That is absolutely amazing. On the self-service side, increased deflection. You heard some statistics this morning I believe, but we are seeing customers get deflection rate improvements of 83%. 83% people are being connected to the information or the work or the service that they need. And that just saves more time. People could be more productive about other things. But let's really talk about what Kush built and how that turns into the experience that drives these numbers. So Brittany is going to drive for me again. Very much appreciated. And what you're looking at is the workspace for the people who are now interacting with this process. Right across the top you're gonna see some chevrons. Customer request, ground crew tasks, customer service tasks. Those are all the things that Kush designed in this process. You also see that when there are exceptions or there are things that are required, that box in yellow, address is outside of supported area. Agents run into this all the time. What do they do? Well, they probably go to a knowledge base, they ask a manager or they do something like that. They don't have to anymore, not with Now Assist. All they have to do is go to the Now Assist panel and they can ask the question, how do I handle situations where delivery address is too far? Okay, we will ask Now Assist and we get the answer. Here is the knowledge base article. Here is the prebuilt exception flow. No muss, no fuss, I could click on the button and send it along. But it's not just that, it's also about the performance of the process itself. Let's say that we've been operating this for about six months at this point. We wanna know how we're doing and whether or not we can do it better. I can go right back into that Now Assist panel and I can tell it to build me a visualization, a dashboard that tracks the performance of this process that I've designed. And just like that, the dashboard was created and I can review it. So go ahead and review the metrics, please. And here you see the things that you would expect in a process, mean time to resolve, cost, customer satisfaction, it's there. And for those of you who have worked with some of our tools like performance analytics, amazing trend performance tool, sometimes it's a little bit difficult to create a KPI. This did it for you and it did it in a matter of seconds. So let's go ahead and open the dashboard and see what it looks like. And look at that, if you remember, Kush was talking about his experience in Amsterdam. They have the lowest mean time to resolve, which is pretty amazing at this point. So clearly he's been doing his work and delivering that experience that he wants. But I can actually create a new visualization very easily. So let's go back to Now Assist. What's the recovery rate over time? Here's my new visualization that's gonna get created. And I can look at more detail if I want to. And we're at 80%. Not bad, we can do better. So we're gonna add it to the dashboard and so we can monitor even more going forward into the future, making sure that we're trending in the right direction. So agent productivity, data, understanding what they need, when they need it, and in this case to be able to drive great consumer experiences. Now let's see what this actually looks like from a phone. So unfortunately, I'm the problem because my luggage was lost. So I can use the Kush world travel app in order to be able to hopefully do something about that. Well the first thing that I'm going to do is I'm gonna log into the app and I'm going to open up the chat bot. And I'm gonna say I am filing, need to file a loss claim. Okay. Now what's happening here is that we're actually connecting to that catalog item that Kush built. And if you remember it had a series of questions that it was going to ask me in the context of my bag. So let me file a lost baggage claim. Lemme make sure that that works. And here are my flights. I am a frequent traveler and I've got a number of itineraries, so I can select the one that I'm currently on because I'm arrived. And again, I'm here, I'm not at home. I can choose my specific bag. I have two checked. I see it because of the knowledge graph. I understand exactly what's going on. Unfortunately, the bag that was lost is the one that had my toiletries. So I am well dressed, but you might not wanna get that close to me right now. What does my bag look like? How many folks here have lost a bag? I should see, okay, that was a reasonable number of hands. That's a good thing. But you have to go through this process. Describe the bag. So I'll describe the bag here. A black away bag, soft. Alright that's been submitted. And it wants to know where I need to go. The resort. Location services right within the mobile app. Very easy for me to submit. And there's my claim ticket but it also knows based upon what Kush has put in there that I might need some additional stuff. And here's the hospitality kit question. Alright, what are my options? Better Beard Deluxe, it's on brand, kind of like that. But lemme check the other options. Shaving kit, you don't know how long it's taken me to grow this. I've had facial hair for (indistinct) years. And then the last one, mani, I could use that but I probably need the beard kit more. So I'm gonna just swipe back here and I'm gonna gonna hit that one right there. Okay, are we delivering that to the same address? Absolutely. Anything else, just let me know. So Brittany, thank you for driving that. And you'll be back a little bit later with Dorit. And look at that, I just got a notification that something is out for delivery. All right. So what you saw there was the demonstration of the agent, the demonstration of the self-service for the end user. And... Well, how is this for service? A nice little bag. I need to tell Kush that he did good here. Oh, and not just beard stuff. I've got, well, skivvies, you don't need to see those. Some socks. And look at that, I think there's even some chocolate in here somewhere, right there. What an amazing experience. And this is absolutely putting AI to work and showcasing the value of the platform and the things that you can do in order to be able to drive outcomes. So when is all this gonna be available? It's gonna be available, again, this year. So be looking for things on the agent side with respect to being able to find those sources of information and being able to generate replies and multi-article Q and A. And making sure that we get to those external systems like you saw Joe demonstrate in Bill's keynote yesterday. And certainly on the experts and the agents, sorry, on self-service. Sorry, self-service is the multi-article Q and A. And experts and agents, recommended reply, navigation and search, all those different kinds of things. What you've seen is what we're delivering in 2024. Now the reason that we're not showing you and saying 2025 is because AI is moving so fast. And what we're doing is we're looking at all the trends and trying to figure out how we do that. And what I'm very happy to be able to do is to introduce somebody to the stage who's going to take you through the AI of tomorrow and the future. So my friend, Vice President of Product Management for Platform and AI Innovation, please welcome to the stage Dorit Zilbershot. (energetic music) - Wow, right? That drone, almost as exciting as our 2024 roadmap. So we've showed you how Now Assist is changing the way we work. As we're looking into 2025 and beyond that, we're looking to push the boundaries of Now Assist. And redefine the way employees, agents, and admins work by introducing new AI capabilities. So in the next few minutes we are going to talk about autonomous AI assistants, AI twins, simulation agents, and local models. I know it means nothing right now, but hopefully by the end of this you understand everything. Remember the days where we had to memorize phone numbers instead of just tapping on a name on our phone? Or when we had to use physical maps to understand how to navigate to a new place, instead of just asking Waze or Google Maps to take us there. That effortless experience is exactly what we wanna bring to the ServiceNow platform. We invested a lot in self-service experience, helping you increase your deflection rate. We introduced question answer capability. We also introduced taking a catalog item form and make it a conversation. But we don't stop there. We wanna introduce autonomous AI assistants to really make self-service experience effortless. Let's see what we have in the plans. So we have Heath, right? You just saw him go off stage. And Heath is going to the employee portal. And luckily, Now Assist knows a lot about Heath. So it knows he's in Las Vegas. And it also knows that Heath is waiting on a very important update from Jon, 'cause Jon just approved giving two VIP tickets to two of the AI keynote attendees to the Pitbull concert on Thursday. Amazing, right? But Heath wants to make sure that those tickets are delivered here. So two lucky winners will get them. So instead of going, finding the request, put in a comment, and hoping everything will work, all Heath has to do right now is just ask Now Assist to do it for him. So you can make sure that the tickets are delivered to Hall A, and that's it. Just like that, Now Assist will update the ticket for him and make sure that the tickets are ready for you. Effortless, right? So you've seen Heath, he lost his luggage. He's been having a really rough day today. So not only that, he also cannot log into the wifi of the event. He's been able to use his hotspot on the phone, but it's so slow. So he wants to get help from Now Assist. So you saw this morning we talked about multi-modals, being able to take a picture and having AI understand. So we can also take a picture of the issue, upload it to Now Assist, and it'll automatically understand exactly what is the issue that he is facing. And not only that, it will find the information that is relevant to solve the problem. But we're not just gonna give Heath a list of steps that he needs to manually go through. We're not just gonna give him an answer. Since Now Assist has access to the knowledge graph and is able to have this autonomous AI assistant, it's able to take the knowledge base, understand exactly which steps are relevant for Heath, and guide him through the process. So it'll ask Heath, are you experiencing this on your MacBook or your phone? Well, Heath is experiencing that on his MacBook. Now it'll ask, how often does it happen? Well, it happens all the time. So now, Now Assist will understand, well, I can't really help. We do need a person to look into this and fix the problem. So it'll ask Heath, do you wanna talk to an agent or do you want me to create an incident? But as you know, Heath is very busy right now. So he will just ask Now Assist, well, create the incident for me. And just like that, an incident is created on Heath's behalf and he can change a priority to make sure it's a high priority. So this issue should be resolved by the time this keynote ends. Simple, right? That's amazing. So if we go back to the slides, we wanna show you how we take this to the next level. Because what I just showed is still a generic way of automating tasks and making work effortless. But what if we can make this our own? What if you could have Now Assist learn how you like to work? What if it can learn from you how to do some of your tasks? With that, we are working on AI Twin. This is an autonomous AI assistant that is yours. It's your personal assistant at work. You can customize it, you can personalize it, you can tell it how you like to work, you can teach it how you work, and you can also have it do some of your work on your behalf. Let's see how that will look like. So I'm an IT service agent. And you can see my beautiful avatar. Looks a lot like me, right? And I can give Now Assist some preferences. So if I open my avatar preferences, I can give them some personal information about me and how I like to do work. I can tell it I really wanna work first on urgent tickets because SLA is important in our company. I also really care about the quality of the service delivery. So I wanna make sure that when Now Assist replies on my behalf, it uses active listening phrases. So that's great, I personalize Now Assist now. So once I save this, you can see that my priority ticket has changed and now, Now Assist prioritize this high priority ticket. So let's take a look at that. So you've got this wifi issue. And Now Assist is not only able to help me prioritize my work, it's also doing a lot of the heavy lifting for me. It's able to go and find the information that will help solve the problem. And not only that, will create an action plan. So it will ask me to review the steps. Let's look at them. Yep, looks good. Let's ask Now Assist to run it. All right, it's run it. Step one, complete, step two. Wait a minute, what just happened? It failed. Okay, so it doesn't know how to do step two, but I know. So let me teach it. I'll give it some instructions of how to do step two. And let's run it again and see if it's able to do it. All right, step two complete. Oh, amazing. So it did all the work for me autonomously. And it not only did that, it actually drafted a really nice message to Heath, making sure Heath knows he is the most important person for me in the company. Amazing. And remember that step that I taught him, it can now automatically create that as a workflow. So it doesn't only help me, it'll help other agents in the work. So Heath is happy backstage, his wifi is now working. I'm happy because that was such an effortless experience having this personal assistant at work for me. So great. So we talked about the request and we talked about the agent. But I know it probably sounds like science fiction, what I just showed. Well luckily, we have our own AI research team, an amazing team that is really looking to push the boundaries of generative AI and help us bring some of these visions into reality. With that, I want to introduce Nicolas Chapados, Head of AI Research, to show you some of the things that his team is working on to make this a reality in 2025. Nicolas? - Thank you Dorit. So yeah, Dorit, those capabilities might come way, way sooner than you think. Our goal at ServiceNow Research is to look just beyond the horizon, a couple years in the future at AI breakthroughs that will power the capabilities that matter to our users. And just to take one example, we started working on large language models for code early in 2022, way before the ChatGPT craze. And that led to the release of the StarCoder model that is today powering our Now Assist for code experience. And what I want to do today is to tell you about work that we started doing last year about autonomous web agents that can do things on your behalf on the web. So in other words, we're teaching AI assistants to use a web browser just like you and I use it, by pointing and clicking on buttons, by filling out forms, by navigating from page to page to gather information and to accomplish actions. And the best part is those autonomous web assistants, we can command them by giving them simple instruction in natural language. So let's see how this will work. Let's show the demo. What we are going to be talking about is how to automate a task that even in the future we still will have to do and that all of us hate doing, filling out our expense report following business travel. So I want to show how I fill, I use my AI assistant to fill my travel report to come to this conference. On the left of the slide, you will see the conversation I'm having in natural language with my AI assistant. I can give it a simple request and on the right you will see the agent in action. All the movement you will see on the right is my assistant using our enterprise tool for expense tracking and using it automatically given my request in natural language. So let's show the demo. I start by giving it a request to fill my expense report and find the two receipts that are located on the desktop. I have an airfare and a hotel receipt. And we see that the assistant is able to upload those receipts into the tool and hit the Create New Report button. I will show the next part of the slide, sorry, the next part of the demo a bit faster. We see the agent in action. If we go to the next part of the demo, we see now the assistant using the tool by looking at each receipt one after the other, fixing the information that has been uploaded incorrectly into each receipt. So field by field, filling out the airfare so it's automatically fixing that information one field after the other. Quickly finishes the airfare and now it switches to the hotel, the same thing, it fixes the hotel information that has been entered incorrectly. And boom, in no time my expense report to come to Knowledge is filled out. So Dorit, I predict that this technology is all going to save us a lot of time. - Amazing, wow. That's great, right? So I was fortunate enough to spend some time with some of you here during Knowledge. And it was great to see your excitement with Now Assist and the understanding of the value it's gonna give to your business. But one of the things that came up again and again in those conversations is the fact that you wanna have guarantees to your investment. You wanna know upfront what will be the business value before you even start the implementation and before making the decision. So lucky for you, one of the things that we are working on is simulation agents. Imagine if you could test a ServiceNow capability in a virtual programming environment where you can actually understand what is the value before you even start your implementation journey. So remember that self-service experience that I showed? What if you could know the deflection rate and what content contributed towards that before you even started thinking about how to turn this on? So let's see that. So we've got DJ. DJ is our ServiceNow agent. As you can see, DJ already has his AI twin. And so Now Assist was able, based on DJ's preferences, to prioritize some optimization opportunities. And so the first one that we're seeing is improving deflection through automation. So Now Assist is offering DJ to run a simulation based on his company's data to understand what will be the deflection rate. So let's do it. It will first ask, do you wanna do it on ITSM data or you wanna do it on HR data? Why choose, let's do both. Then it will ask about the seed data. How far back do we wanna go to understand what do employees ask about? So we'll go with six months 'cause that's what the system is recommending. Looks enough data to understand how employees are interacting. And now the simulation starts. Step one, the system will assess all the existing conversations. So we've got about 73,000 created in the last six months for IT and HR. Now it understands exactly what type of questions employee asks. So once we have that information, we can go and we can generate conversations automatically. We can generate 250,000 conversation. Imagine how long that will take you to do that inside your organization. Next step will be to run a simulation. Let's run these 250,000 conversations and understand what will be the user experience? And so just like that we can see, without doing anything, we didn't turn on anything in our physical environment, we can see that the deflection rate is 56%. And not only that, Now Assist is also creating us an action plan to get to 70%. Remember that catalog item that Kush created at the beginning? What if Now Assist was able to tell him which catalog item to create to save that step as well? So that's great, let's apply those changes and let's see what are the new numbers. And so we see 70% deflection rate and 353 hours saved. That's great. So now let's ask Now Assist to just deploy it in a sub-prod environment so that the business users can actually test and validate. This will save you and DJ months of work. Now Nicolas, this is a hard question. How real is this? - Well, it's way more real than you think, Dorit. Actually by building on the autonomous agent, autonomous assistant technology that I described earlier, we can build simulation agents right inside the ServiceNow platform that are digital replicas of real users on the platform. So each of those simulation agent has a persona and has a task that they are looking to accomplish on the platform, just like we are when we are using the ServiceNow platform. And by tracking what each of those simulation agent is doing, which KB article are they consulting? Which catalog item are they trying to use? We can look at where they are getting stuck, what information is useful to them, or how many tickets do they have to open because they cannot find what they need? By tracking thousands and thousands of those simulation agents, we can predict the deflection rate of your current deployment of the Now platform. But importantly, we can tell you what are the missing KB articles, what are the missing catalog items that if they were implemented would make your users even more productive on the Now platform? - That is amazing. Now Nicolas, I have to ask you one more question. So we talked at the beginning about four things, but we only mentioned three. Anything else exciting AI Research team is working on that we wanna share with the audience? - You bet, Dorit. So I want to talk about local models, super exciting thing. And it starts with the belief that the future will be hyper-personalized. In the future, each of us will have our own generative AI model that we will want to run on our own devices. And this will open the door to fantastic conversational experiences that will be private, that will be secure. And importantly, that won't require a huge data center located halfway across the globe to support. I have a live demo. Do you want to see it? - Yeah, let's do it. - So here, what I want to show is a conversational experience that is running 100% on a laptop that is not connected to the internet. The wifi is turned off. Normally those experiences require huge data center to support. But today we're going to show you something that is running locally on this laptop. Now, when Heath was having his wifi problems a couple weeks ago, I was in the same office as him and was trying to get my travel to Knowledge '24 approved by my manager. So let's see how this went. So I'm going to ask my bot here, I want my travel approved. Okay, where are we traveling? Well, of course we're coming here to Vegas. Okay, the destination is Las Vegas. What's the purpose of the trip? Well, of course I'm attending the Knowledge '24 Conference. And I want to go from May 8th until 13th. Okay, so the system picked up the dates. All right, but you know, I changed my mind, I want to do a little bit of sightseeing, so I'll ask the system to come a day earlier. Let's see. No problem, it picked it up. Okay, such a great system. Okay, I'm going to go ahead and book this request for information. Okay, so the system is not connected to the internet, so it will book my travel whenever it comes back. But Dorit, it gives you a sense for the kind of great conversational experiences we can expect in our hyper-personalized future. - Just amazing, Nicolas, thank you so much. And thank you Brittany for running all the demos. So to summarize, we showed you some great experiences of new AI capabilities we wanna introduce. We showed you autonomous AI assistants that will make work effortless. We showed you My AI Twin that will become your personal assistants. Simulation agents to make sure that you can understand the ROI before even starting the work. And then local models to ensure privacy and make sure that you can book your travel without wifi. Now, as you've seen here in ServiceNow, we're not just taking incremental steps towards the future. We are creating the future. And with that, let me bring Jon back on stage. - Thank you, thank you. No, I want you to stay with me. Oh, okay. - So we're gonna close. Don't leave. Trust me, don't leave. We're gonna close with Pitbull on stage. No, we're not gonna close with Pitbull on stage. However, we do have VIP passes for somebody, some lucky person, underneath your chair. So please holler and scream. And if you don't find one, it's under a chair that somebody has already vacated. Oh, we got one. Woo-hoo! - Nice, congratulations. Hey, I wanna thank everybody for your time. - And there's another one. We ran a little long. Please go next door and see the show floor, and we will see you next year. Thank you so much. - Enjoy.