(upbeat music) - Hello, thank you so much for joining us. Welcome to another Now on Now webinar where we are going to be talking about how we're using GenAI meet Now Assist to drive more productivity at ServiceNow. So we all know Gene AI is a very, very hot topic. My name is Bree Wilson. I'm a senior executive writer on the Now On Now team. And we're telling tons and tons of stories about how we're using our own Now Assist at ServiceNow to really experiment with Gene AI and drive value across the organization. I am joined by Rajeev Sethi, our group vice president of Emerging Technologies at ServiceNow, and he is going to share our story about how we are customer zero and we're really leading the charge with GenAI and using our own platform and technologies. Rajeev, take it away. - Thank you Bree, really appreciate it and welcome everyone to this webinar. Good morning, good evening, wherever you are located. Happy to share our journey at ServiceNow being customer zero where we are imagining how the work is being done and then being now and now we want to make sure that we are using our platform every day to drive the speed and the scale across the enterprise. Alright, so talking about GenerativeAI and as I talked about being Now on Now being a customer zero, we are making sure that we can leverage our platform to the best extent possible. Being customer zero, everything which comes from our product team. We are the ones who are deploying it, making sure it works really well before it gets to you. And then making sure that we get value out of our investments we are making in our product and our platform. So that's part of our role over here at ServiceNow. So I had the privilege and the opportunity to travel across the world over the last six, seven months meeting a lot of customers and what we've been hearing, especially from the c-suite over the, the priorities for 2024 has been how's the technology landscape changing? It's changing from the cloud to different, different aspects in into the infrastructure side of it. They want to see how the employees, the impact to the employees having, whether they were returning back to the work or in some other particular fashion, how that is going to help our employees to really engage more and more into the organization. They want to see how customer experiences are getting better and better because that's an important part. How we going to increase the developer productivity? And then generative AI is top of mind of every c-suite, every CEO, every CFO, every CIO. And this is a pretty common theme across the globe. As I said, I had the opportunity to travel and meet a lot of senior leaders out there from different verticals, whether healthcare from finance, from retail, and across the spectrum of all the different industries. So this is, I have been on top of mind. And then when we talk about everybody's business objectives, everybody wants to grow their customers and everybody's looking to make their employees very productive and what we see is that AI is one of the ways to get there. It's not the only way, but AI is a game changer out there, especially in last one year, one plus year, how GenerativeAI has make a major, made a huge impact and a move that is absolutely we can see made some great strides already in terms of creating productivity and which we'll share with you some of them today. So when we look at AI, we look at AI from three different dimensions. We are looking AI because you don't want to use AI for the sake of AI. We see that AI can improve your experience. So what you see on the screen, we are very familiar with in our consumer life outside our work where AI is helping us to create some great experiences out there, which we are very familiar with it whether especially a lot of social media companies out there is using AI to provide you the recommendation or you go to a shopping site or any other commercial consumer facing sites out there. So that is where AI is being used a lot. And AI is used also behind the screens where it helps you to make things move faster. So how are we routing things? How are we moving things from one state to another state? Like are we passing the information to the right people at the right time? And the third place where AI can help you is making decisions. As we all know, there's a lot of data out there and you have to go through it and figure out what is the next best action you need to take. So the decisions you are making. So when when you look at it AI, you need to look, think about it, how is it going to help you to drive better experiences, how it's going to make things faster and can it help you to make decisions and take the decisions at the right time. And that's the lens we are applying across the board. So our vision is evident ServiceNow when we look at it that AI is everywhere. Every part of the ServiceNow every department is using in and it's part of everybody's work and that's what we feel is going to transfer your experience, it's gonna help you to make better and faster decision. Now how that will be done, we feel like every employee has an AI powered assistant. We call it Now Assistant ServiceNow. You have co-pilots, companions, there are a lot of different ones which we know commercially available. And then there are a lot of going to be homegrown, those AI powered assistance which will help everybody do the do their work. So it's gonna be based on the person. If you're a developer, if you're an agent, if you are somebody in marketing, you're in sales. Every persona will get, will have an AI powered AI assistant to do their work. That's what we see our vision is today. So as you take it a step forward, our AI, AI has not happened overnight. AI for, at least at ServiceNow we have seen the AI maturity journey has grown over last seven, eight years. We started in the foundation area where we took, it started in the analytics area as we are very familiar with it, analytics deals with a lot of data and data is the foundation for a lot of AI. So we started looking in finance and sales for reporting point of view and bringing all the data into a consolidated platform from multiple different sources. Then in in parallel we started looking at real time engagement, whether it was search, which we are very familiar outside our work, but inside the work search plays an also an important role in the enterprise and as part of virtual agent conversations and configurations out there. Real time reporting and things like that. So that's where we started use of the real, getting real time information from the data which was put in the foundation. Then we started looking at prediction. Prediction is like can we predict, can we predict which customers are going to renew? Can we predict as we get leads from it, how do you score the lead so you can focus on the most important lead out for you? Can we focus on some of the customer, how well they are using our product? So that is where the prediction came into play. And then we started using prediction around sales, marketing and finance. Then we went into the prescriptive mode. Prescriptive is where we are providing our recommendation. What is the next action you need to take or what is the next training you need to take? So things like that. And it was incorporated in the workflows for the user out there. And then came generative AI last year. So generative AI now and integrated with search and virtual agent is providing based on your questions answers out there and solutions out there to generate content, to generate email. So there are those kind of outputs we are getting from generative. So that has been our maturity journey in different phases within ServiceNow. And when we look across from our ServiceNow platform point of view within ServiceNow, as I said internally we were on that journey but from a platform point of view, we have been making investments over year, over year, over last six, seven years. Starting in 2017 we started making investments into the prediction side classifications of the content there. Then we went into the chatbot, virtual agent search and LQ and on and on. And if you look at it every year more and more AI capabilities have been in the platform and with every release you, as a customer have been benefiting from it. As recently as last 2003 in the Vancouver release we brought in some a lot of generative AI capabilities, which was just a starting point. And in 2004, which is this year as part of our (indistinct) release, we have already released some really good capabilities out there and it's just in the first four months of the year and you can imagine what more is going to come for the rest of the year in this space. So really, really excited about within the ServiceNow platform what is coming. And we all know ServiceNow platform as AI is embedded, but it is built on workflows, it's built on a single data architecture, a single data model, and you have the technology workflows, the employee workflows, the customer workflows and AI is embedded in the platform. So you're not going anywhere outside the platform to leverage AI which is all the capabilities which is there, which is search, conversational intelligence, whether it is prediction, operational intelligence, document intelligence, developer. So all this AI and GenAI's capabilities all embedded in the platform. And when you combine that with that, we look at it that when you take the unstructured data, which is part of generative AI and when you look at all the automation which is available in the platform, which is all around structured data and when you bring the structure data into the workflows, those workflows become supercharged. So for just example, an agent is working on a ticket, now those workflows are supercharged with the data, unstructured data like solutions out there and things like that. So that's what we feel really, really strongly about that all the workflows which already exist in the platform are the ones you are creating, which you can easily embed a lot of the unstructured data from GenAI and help you to supercharge your workflows out there. So as I said, our platform where you have so many of different workflows already available to you, whether in the technology workflow and I'm sure you're very familiar with some of these products or the suites we already have like ITSM and ITOM and SPM or on the employee workflow like HRSD on the or the CSM or our creative workflow like App Engine. And for last couple of years we have been investing a lot in the finance and supply work chain workflows and we have a lot of industry specific workflows also available in our platform. So just imagine all the AI capabilities which is there and when you apply to the platform with all the intelligence which is there and with all the prebuilt workflows which are available, this accelerates the value to you as a customer and for us definitely. (audio garbles) So let's, everybody talks about in the generative AI area, what about large language model? And what we have done in within ServiceNow is when you take the our platform, put the large language model next to it, which we now are shipping to our customers and we are leveraging it internally and with now it says which is where how the users engage. We see this a combination of these three things which makes the Now Platform the AI platform. And I'll show you some examples that show you some demos of that today. And when we talk about large language models, we look at it in two different places, we look at two different approaches. One is on the right hand side what you see is the domain specific large language models which we are shipping to the customers out there. So you don't have to worry about the data we are, those are very domain specific whether you are summarizing, whether you're generating code or generating knowledge articles, those are very domain specific, they're trained on our data or it is trained on customers who have opted in. So don't worry about it, we are not just taking your data and training this model but they're domain specific. So there is a very less chance of hellenization or getting you incurrent results because it is meant for that particular purpose only. And then we are retaining within the ServiceNow so your data doesn't leave out anywhere and then you have general purpose more models out there which can do a lot of broad intent stuff out there. It can respond to multiple topics, things like that. But in that case you are sending the data out outside of ServiceNow and by the way we give you both the options. You can use pre-built workflows integrated with larger language models sitting next to our your ServiceNow instance. Or you can use our generic controller and connect to any of our, any of your own large language model which are internal purpose or you may have tuned it and used it for your internal purposes and you want to integrate it, you can do that. So if we don't stop you that and that's what we have done internally. So we have in fact domain specific models for our use cases plus we have using some general purposes models where it's not related to ServiceNow platform especially in the sales marketing area, right? So when we come to the next part of it, we see everything in, for me part of the IT organization, we call it DT, I'm like many of you on the call today. Very important for me it is that how much of investments I need to take, how, how I'm gonna deal with the infrastructure, how I'm going to deal with the security because we all know security, privacy, legal are critical parts of discussion in the AI realm today. So in this case you are not creating, I didn't have to create any new infrastructure, I didn't have to make any security changes or bring in any large language models. Obviously I had to go through and make sure our security privacy and legal team reviews it and addresses any address, any risk related to it. There was no development work, I don't have any out of the box. A lot of these features work very seamlessly in a very, very quick time. It's like less than 30 days you can up and running summarization and generation in a lot of capabilities out there and you don't have to move any data, we don't have to move any data out of ServiceNow instance into some other place to make this thing work. So it was, that's reason also the speed helps it because you know when the data starts moving, you start getting a lot of investments goes in that. Okay and then lastly, I don't have to develop any new skills. There's no machine learning skills, there is nothing like that with platform expertise which I have within my team, I was able to very quickly start using the capabilities and then expanding it on day-to-day basis out there. So that's what I, at least I could benefit from leveraging everything in the platform. And now as I mentioned it is the, it's where how your user engages it, it brings the intelligence into every part of the business whether like cases, whether you are having a conversation through virtual agent, whether you are creating content, whether it is like knowledge articles or we call it text to X. So whether it's text to code, text to flow or text to app or text to catalog. So a lot of those capabilities to help you to really make sure that your employees are productive, you can create your great experiences like conversations through our virtual agent and quickly think move the platform out there. So that's how we look at it. When we looked at the entire approach of to generative AI, we wanted to make sure like you all are that it is, you can trust it and that human is in the loop. So we have applied the same principles that every AI analysis generated output is watermarked with saying that this is generated by AI. So the user know the same day they are expected to validate before they consume it as part of the work. And we are not like creating generative AI solutions everywhere. We are leveraging in platform like within ServiceNow platform we have Now Assist, we also are a Microsoft customer. So we are looking at Copilot, we are a Zoom customer so we are looking at Companion. So we are looking across the board with, for all the different applications we have, what kind of capabilities they're providing and we are looking from the same lens, hey do I need to move my data? Do I need to deploy any new infrastructure? How's the security privacy of it? Do I need any special skills to maintain? Same thing applies to everybody and that's what we are doing it for ourselves too. In some cases I said we have to invest in large language model where it makes sense to build some solutions which are not coming from any of these providers. So around that we are putting logging services, moderation services, moderation is what kind of data somebody is sending, what data is coming back, like are you sending things which you're not supposed to send. So we can moderate that feedback from the users, how we test it based on all the different capabilities. So that is the approach we are taking. And then as you shared in the opening, you are experimenting, we did a lot of experimentation and then we deployed quickly. Today we have more than like 20 use cases in production and another about 14 of them currently worked in process. So we are very quickly iterating because we want to see what kind of business value we can generate from it. And we know AI, AI takes a while before you can really see the the real use of it and how it driving value because the models have to get better output has to be better, users have to get comfortable with it. So we have to keep on driving adoption, learn from the feedback we are getting. So that's the approach we are As I said, and we are making sure we are infusing our own products inside the workflows and inside our work we are doing, whether it is on the customer side, on the HR side, whether it's on the IT side, operation side and then the Now Assistant search and virtual agent are pretty, pretty robust out there. We are already seeing a lot of values in that sense. So continuing with the journey, let me now as I talked about to you, gave you some background about our approach, talked about how our platform we are leveraging internally. Let me show you a couple of demos and use cases so that you get an idea about how we are realizing values inside it, inside ServiceNow. So begin with the first one is how we are looking at sale service. So sale service is where we are going to demonstrate how a user can get summarized results, so then they don't open an issue. Here is an landing page of our employee portal, we call it My Service Now. User comes in, generally they're coming here to report an issue. They or they're trying to search something when they report an issue. Generally they're like, and it doesn't matter whether they have an IT issue or an HR or they have a question out there, they can type in, like unable to log into Mac out there. So we don't immediately give them the option to submit a ticket, but we ask them like hey do you want to see if I need any help here? So when they click on continue, they they get the, they get now a summarized answer when they're reporting an issue. So like how I'm able to unable to log into a map earlier you would've got a like a link to a knowledge article, you have to open and read it in this case now we give them a summarized generative AI. As I said this is an AI generated response answer generated by analysis. So you can see we are making sure the user is aware of that where these answers are coming and they can also see the source where the, which knowledge article is used to summarize it. So if you look at it very quickly, they can get these steps as they're trying to open an issue and obviously if they don't like this answer they can give a feedback of thumbs down and then they can go and open an issue out there. So this has resulted for us a 54% deflection resulting from these answer generated by AI. And you can just imagine how huge it is. We get close to about 10,000. We are at 20,000 employee 2020, 2,000 employee company and ServiceNow and we get to close to about 10,000 interactions or incidents internally users goes and create them or the request and then more than 54% of them are getting now deflected. That's huge amount. You can just imagine, users can solve things on themselves and then they're not opening the tickets out there. So again, these are some early GenAI like we just recently launched it and we are seeing very positive results out there. Okay? And then now we saw from an employee point of view, now what happens? The employee still doesn't get an answer, they like it or it's not there and they open a ticket. Typically what will happen is it'll go to a support agent. And now by the way what I just shared with you, it's also a similar experiences there on now support, which is our customer support facing side. And by the way, you all are customers hoping that you are all customers today who are logged in here and you are familiar with the Now Support. And if you go to Now Support and try to open a case there, you'll get a similar experience of of getting a result back based on the issue you're trying to or case you're trying to create like relate to your instance. And by the way, if you go and now support and use search, you'll also get a GenAI result. So these are a couple of places where we have been using generative AI to come up with the answers rather than give you a knowledge article. So when we talk about empowering agents with generative AI, so as I said, if the incident gets still opened out there and if it's an incident which has been opened and you can just imagine when you can keep on creating work notes, that's information exchange between the agent and the user or they can create some private notes, after a while, after two days, three days if you are reassigning that ticket generally sometimes ticket get reassigned or you go back and open a ticket and sometimes just imagine in you are as a customer, if you call help desk anywhere and the operator or the agent will say, Rajeev, can you hold for a couple of minutes? And what, imagine what they're doing, what they're doing is they're reading through the work note, the history, what has happened before they want to do it. Now instead of waiting for two minutes or three minutes for that user to, for that agent to read it, you can summarize, you can summarize that incident pretty quickly for them. And when you hit the summarize button you can see they can get a summarized result within seconds of everything which has taken place. Okay and what, what was the issue the user reported and what are the key actions were taken? So you don't have to go, the agent doesn't have to go read notes of over and over again to figure out what it happened, what action we taken because you don't want to tell the customer to take same actions which some other agent might have helped in try to do it to resolve that issue out there. So summarize is a pretty key capability to get started very quickly in that and since we have launched it, and this is really amazing if you can look at it across the board, summarization is one of the things, but generating resolution note, generating knowledge articles we are seeing between IT agents and customer agents which are customer facing like you are, IT agents are facing internally to our ServiceNow employees. We are seeing about 50 full-time equivalent productivity announcement. Again, by the way I wanna qualify this is annualized, it is, we started in January, in the first three months we saw the results and we feel very really confident that we can, over the year if we continue with the pace we are, we are going to see 50 full-time employee worth of productivity which we will be able to save for Service Now just based on this data out there. Okay, so that's pretty significant I can imagine. So you saw the employees or customers opening incidents or cases, you saw the agent. Now let's look at how we are helping our developers. So from a developer point of view, you have heard about text-to-code, I'll show you an example of text to workflow. As you know, ServiceNow platform, you create workflow and to create a workflow you have to go and drag and drop everything like you want to create the shell out there. So in this example is that now if you want to find all the newly created records in past one day and if they were not assigned to an incident to level one group and then send it a notification to that group. Simple English you time pin, this is your prompt, you write it and you say build with Now Assist. And when you build with Now Assist what it does it, it'll create you, if you can see now it has created this look up the record list the item if these conditions are met and based on that, update the record and send the notification. So it creates a shell for you, and it significantly reduces the time for anybody who's trying to create a workflow out there. So you don't have to, this would would have been been the landing page for the any, anybody who's trying to create a workflow and try to bring in one field at a time or one of these drag and drop interactions and then build it and with now by just typing in simple English, you are able to get a good, decent workflow already created and then you can go and tweak it and deploy that. So that is how we are increasing the whole developer productivity out there and we and we know that initial, everybody thinks about it, hey are developers accepting that code? And one of the things which we have started tracking is that how much of acceptance of the generative AI code or any kind of output any users are getting and whether they're liking or not. And in this case more than 50% of the users are accepting the generative AI code out there, which is significant if you can see it, especially when you have to write complex workflows. It is not gonna be easy and obviously the easy ones are relatively good. So when you average it out, more than 50% of the code is accepted by our developers out there. So when you look at all these three use cases, I just wanted to give you a a overview of that and I said we have a lot of different examples out there, but across the board, whether you are a employer or a customer, you can have conversations like virtual agent is a pretty cool one to experiment with that. You can integrate very easily with external systems or we are on our agent. So not only when you are creating, summarizing, you are creating resolutions node, you are also generating K articles and managing all the interaction. Like if you're having a chat interaction with your end user, you can summarize that and I said on the developer side we have deployed a lot of different use cases which is increasing the developer productivity and just getting more value out of the platform for our developers out there. So here are some of the use cases. Just want to share on a higher level the first vertical, the go to market, this is where it's outside a platform but we have integrated something like Sales Assist, it's like a ChatGPT kind of an interface which we have created for our sellers where they can go ask a question about a product or a solution and get the answers. We are doing a lot of other generation like generating RFPs and emails which go out to our customers and our prospects out there. But when it comes to our platform, this is where a lot of capabilities exist in the platform. Like as I said, search virtual agent conversations are now you, once you have a catalog and catalog has say four fields on it, you can make that conversational by a click of a button and you don't need to now go back and rebuild the conversation in virtual agent and when the catalog changes, conversation automatically gets updated and your users are grading a true experience of a conversation rather than a catalog getting open entering four fields, now that is more of a Q&A out there. Incident deflection, I just showed you one. And the same thing on the customer side. When we talk about internally to what we call the global business services, which is like IT, HR, legal, finance or facilities chat summarization, case summarization, which I showed you an example, agents can get recommended solutions. So when they are trying to resolve an issue for a user, they can get a solution based on all the knowledge article or similar incidents you have that they can generate notes and they can also generate KB articles. And this is where we see the fastest way for anybody to get started because for chat summarization and for case summarization, you already have the content there. You're not generating fresh content. In fact you are taking words note and summarizing and then based on that you're generating resolution notes. And once the case or incident is resolved, you generate the KP articles and then the KP articles is feeding your search and what we call for deflection in your customer side. So, that within 30 days, especially with summarization generating solutions and KP generation and then start seeing the value very quickly. Now when you come on the developer side, and by the way I just wanna make sure that you are aware a lot of capabilities we have built ourselves internally ServiceNow IT on a platform using a GenAI controller which gives you access to using any large language model if you want to. And that was the beauty of it. So we could very quickly, instead of waiting for our product to come up with a lot of capabilities, we were able to accelerate our GenAI journey using our platform and the use cases we need to solve for. So code comment, text to code, code comment, commenting code explain is a reverse way. We have a code somebody wrote for, somebody wrote it. Now you want to see it in a simple English language what this code does. Very heavily used internally by ServiceNow IT. The refactoring the code based on some best practices. Unit case generation developers don't like to write test cases. So now we have given them capability to do generate cases, functional test cases generation, when you have written a story, which we have also now using generative AI to create stories and acceptance criteria. And now based on that you can write functional test cases. So we are seeing tremendous, tremendous adoption across the board of these use cases as that is embedded and they are our developers and our agents and our employees are definitely seeing a lot of value in that. So Europe, pretty good set of use cases to, up and rolling in like less than six months. And we have as I said, another 12 to 14 use cases which are in works, combination coming from our product and then obviously we extending and building based on that. And what we see now people see what is the value, right? So within first one 20 days of our going live we've been able to get about 5X of our investments back already in the ROI and then we are projecting or targeting that by end of 2024 all the investments we are making in this space, we should be able to easily get 20 times return based on the investments out there. And I know there is a lot of things, whether it is you are, you are investing in licenses, resources, things like that, but a lot of out of the back box capability, as I said, you don't need any specialized skills to implement and run with it. And I hope I've given you enough information here to say about our journey that how we are leveraging AI and making our employees more productive across the board, whether they are employees or customers, whether they're agents or our developers. And we are putting AI to work, whether they are analytical AI or whether it is generative AI. And as I said, our journey has just started. We have been very fortunate to rapidly bring it to the state combination of the work we have put internally and what is coming from our product. And hopefully you can see also get some ideas about how can use get started pretty quickly because we feel that we all can gain collectively a lot and learn from each other a lot here. Please scan the QR code on the screen to view other sessions presented at Knowledge and learn more about we're using ServiceNow technology across the organization to run our own business more effectively. Thank you. (bright music)