Video: Scale ServiceNow AI Agents to Complete Work Autonomously | Duration: 2992s | Summary: Scale ServiceNow AI Agents to Complete Work Autonomously | Chapters: Webinar Introduction (4.16s), Webinar Introduction (109.98s), AI Inflection Point (203.215s), AI Execution Phase (359.79s), Autonomous Workforce Implementation (529.44s), AI Agents Evolution (609.77s), AI Specialist Architecture (823.875s), AI Control Tower (992.315s), Autonomous Workforce Architecture (1128.495s), Live System Walkthrough (1339.21s), AI Agent Resolution (1853.315s), Service Desk Workflow (2063.66s), Roadmap & Next Steps (2271.02s), Q&A Session (2445.515s), Auto AI Experience (2562.95s), Q&A Session (2678.135s), Q&A Session (2810.005s), Closing Remarks (2950.59s)
Transcript for "Scale ServiceNow AI Agents to Complete Work Autonomously": Hello, everyone. I hope you are all doing well. Thank you all for joining us today. So welcome or welcome back to another Now Ben webinar. Some light housekeeping before we kick things off. You are in listen only mode today. Please do use the chat to interact with us. Let us know where you're tuning in from. It's always nice to see how global our audience is. The recording for this session, it will be recorded, and we will be sharing that in our follow-up email in the next couple of days. So, keep an eye on your inbox in case you want to revisit any of the points that we make today. And there will also be a q and a at the end, but feel free to share your questions throughout. And you can do so just by navigating to the q and a tab at the top. Hi, Theresa. Nice. Virginia Beach. Lovely. Cool. So today's session, we will be looking at ServiceNow AI agents. So I'm sure you've all been following the announcements from Knowledge twenty six. So it probably won't come as a surprise that enterprise AI is entering a new phase. You know, organizations are experimenting with AI agents that can operate autonomously, resolve issues, and deliver measurable outcomes. So the key question now is how we can get started with the AI agents to scale our operations. So without further ado, I'm delighted to welcome to the stage our speaker for for today's session, Diego Beltram. So Diego is The UK and Ireland ServiceNow presales director at NTT Data Business Solutions, and he has a wealth of experience within the ServiceNow ecosystem. So you're definitely in good hands. And I will pull down my screen share and, hand over to you, Diego. Okay. Thank you so much, Emma. I'm happy to be here. Welcome, everyone. Again, please use the chat just to share things, and we have a q and a session, at the end. So if we you have some question and you need some answer from me, please use this space, but you can interact with the chat also. Okay? Well, I will start, just sharing my screen here with the agenda, and we just go through this. Okay. Now you are seeing here, we structured this webinar. It's one hour, and we are talking about, starting with the AI inflection point, what we are living and seeing in 2026. It's a new era. Okay? Then we go to specifically the subject for today, the AI specialists, and comparing with what is changing with the AI assistance that we worked with the last years. Then we will define, and share how this is architectured, how the autonomous workforce works in the platform, then go for a live demo, finalizing with the road map what is next and q and a session. Okay? Well, started this last week. We had very good, presentations, a lot of presentations, in knowledge, 26 in Vegas from ServiceNow, partners like Nvidia, we had a lot of information, where this coming from. But, also, from outside ServiceNow, we are viewing this. So new announcements every single day from Anthropic, from OpenAI, from Google, from Microsoft, from Amazon. This just to mention the top, leaders from market, but also a lot of niche applications and AI use cases. Great, actually, during the last, I don't know, four, five years now that it started to be, more common and exponentially, more precise, more accurate, and, actually, more useful for all of us. But we are seeing an inflection point this year specifically. Okay? It started a bit last year, '25, but we are seeing this. We are have passed it by a pilot first AI, and the customer's asking, for example, our employees and and companies working. Maybe this could work. Maybe let's test this AI case. I will see. And the question is not, that. Now we are going to this AI use case needs to, give me some value. So which is the outcome of having an AI solution? So it's not will work, but it changes to which is the value and when the value is coming to the company. So this is an inflection point, and, we get a lot of expectations, over the AI solutions here and there. And in service and world, thinking and autonomous workforce is one of the key points, key elements that replace a pilot to the job is actually get done. Okay? We have this, statement from Amy Zavery. He's president, CPO, and COO from ServiceNow. And last week in Vegas, we had similar statements, from Bill McDermott, from Amy, and a lot of people from from ServiceNow itself. And the message is very clear. Right? Businesses don't need small pilots or testing or promises. All of us, we are looking for solutions, and solution could be AI, but solutions that gets the work done. Okay? And this is anywhere. Why? Because it's not only, the predictable work, and then we move it to something that could be unpredictable, insecure, but need be tested to the current situation, again, which is the outcome of the use case of the the usage of some technology. And this outcome needs to be predictable, needs to be, effective, okay, and needs to have some value for the company that is investing in this new technology. And the reason behind this was structural. So the generic AI, was just like probabilistic generation of could be text, could be image. But during time, this just got better better, And now, they they they are actually pretty good if you ask, if you compare, for example, first initials, Copilot interactions in Microsoft, or for the first questions and answers we received in the early AI, tools that got popular. And this just got more common to to receive, but they don't execute works. So the generic models, just, give us information, data, procedures, data. Okay? But the thing is they don't execute the work. Now we change it. We have some specific agents that actually go to the, book, system and make some reservations, confirms data, send emails, move files, create files. So we are moving to the execution phase. Okay? And the autonomous workforce, in ServiceNow, is actually something that ServiceNow had implemented internally, and they are sharing their the metrics internally. And it's already resolving over 90% of the employee requests. Okay? And the velocity is almost 99 faster than human agents. So this is coming from ServiceNow itself. Internally, they have thousands of employees that already are using autonomous workforce, that is called the digital labor. Right? So, there are agents resolving, the requests. And this is the case we are discussing today, and you will see the road map. This is very accelerated and broad in coverage in different areas, not only IT. And this brings to this message the comparison between, AI assistance with the new AI specialist, okay, when we are moving from generated suggestions to actually full execution of work. And the conversion is, I I try to keep this clear and simple for you to understand. The assistance work it and evolve it actually pretty well from some task level, but they depends on the prompt. It depends on how do you ask this. If I make a sentence, okay, create an image of a a ball, could be a a ball, but, I need to ask to be a red ball. I need to ask to be in a football field, I need to ask. So it depends and relies on the quality of the prompt. Okay? Then the second, it suggests. So if I have some question, can I going to the to the, world of the work in the companies, can I have some, I don't know, holidays, in next May? Maybe you have you have this question, and the generative, AI assistant maybe could say, okay. This is the policy. Look on some knowledge article, generate an answer for that. But, yeah, this is the policy. If you have some days, available, it's okay. If not, just you need to check. It's just a procedure. Okay? So make the suggestions. If we have some IT problem, look for a knowledge base, look for some content, and give me some suggestion again. So the execution phase steps is is actually missing. Then we go to the enterprise context permissions that the AI assistance is not aware of the full context, who I am, where, I am working from. So we have different type of context and permissions to execute, to go to a database, to a system, to connect, to interact, and execute some work. Right? So we go to this and compare them with the new AI specialist where we have an AI with a defined role. So we have an AI agent specifically to support me on IT subjects or to support me, on HR subjects or to guide me through an, project management roles or maybe updates. So the AI specialist is with a defined role, very concentrated, very niche. It's goes autonomous, and autonomous is it has permissions to go think, and it's actually we have the chain of thought, with some reasoning about some task and then decided the best task that suites for this purpose and then end to end execution. So it this is really, really, changing the outcome that we got from, suggestion to actually execute some work. Okay? We will see in the architecture the definition, how it's prescheduled, and how the define the the roles and scope. The AI specialist came native with the enterprise context and the required permissions. So maybe you are aware, that ServiceNow had acquired Moveworks, one of the biggest acquisitions in the history of ServiceNow. It was completed at the end of last year. Now they had announced it that it's ready, and it's live, and context is one of the best part that AI is taking advantage. And the Moveworks product, infused it into the AI platform that, that is ServiceNow, is actually improved a lot. And yep. And in star inside the context, we also have the permissions. So connecting to systems, executing tasks with the right level of permission. So we have guardrails, and we have security, definitions, is covered. Okay? So you can rely on the guardrails. You can rely on how the the systems, databases, interfaces are protected. And the other thing is, okay, comparing with this line that this the assistants are stateless. So you have a long term memory you can set, but between sessions or new sessions can forget some things and you need to repeat again in within the the the context on your prompt. Compared with the AI specialist that is actually, perceiving states, some data and context and learning with the loops of executions. Okay? So it's actually an a learning curve and improving over time. Then we got the assistance just disconnected, from your workflows that you have in the platform, your CMDB, the the items, service definition. If you are having, for example, a a service and you have connected dependencies on that in CMDB, the AI specialist can take advantage of this. Okay? And it's also connected to workflow studio. So if you have some workflows defined, if you have integrations defined, the AI specialist could take advantage of this. Okay? And the last one is more about governance and the possibility to be, a to have an auditory just checking, if this is compliant, if I have some risks, ServiceNow is offering. And no matter which is the product now because it's changed it, it was the the last announcement in April from ServiceNow, and in knowledge, it was reinforced. You have AI at all levels of the platform. So it's before, you depends on, do I have a pro plus license? Do we have AI control tower? And you need to ask for that. Now it's infused in the whole platform, so every action with AI control tower, is covered. You have the visibility of executions, where, when, how, why an action was taken by an AI specialist, and then also connected to the value that it's offering. So is this changing in some way, our KPIs? Is this giving me actually more time free for my team, make some something different? It's compliant. We have the in Europe, we have the EU AI act, and we have some, rules that is that are in place and new rules that will come in place very near term. So we need this, and AI control tower is covering. So we can see very clear the shift from assistance, and they are great, for the purpose, but they are lacking, this execution and autonomous execution. Okay? Again, if you have some question, please use the, the q and a session. At the end, I will cover some some questions that we you have. Now getting to the the details here with the architecture. So autonomous workforce, is actually an specific AA agent that is defined, actually, a group of AA agents because we already have, from now assist, the capability to define individual AA agents and Agencies workflows. You have this capability still. Yeah. It's not replacing, but ServiceNow is adding an autonomous digital workforce, with defining roles, defining permissions, defining integrations, and defining governance. Okay? Considering this, the autonomous workflows is, individual sets of could be digital employees. Okay? Every action again is governed. Every action they make decisions, executions are traceable, and you have all these preapproved through the governance framework. Okay? The Autonomous Workflows relies on the platform itself, so has access to the workflows that are created. You have some out of the box, but if your company has specific workflows, the autonomous agent can just, consider and use it. It relies on the AI components. So, again, the AI agents that you have, the Agencies workflows that are defined. If you are using a language model for some reason and or you have, using a different language model for any reason or an internal, ServiceNow, just keep the doors open. You can use any AI, so it's connected to any AI model. Okay? And then you have the data. And you can see that data is the key element foundational level because the data will be consumed, will in the it depends on the data that you are consuming, the outcomes that that you are having. So historical data, knowledge graph, integrations that you have, zero copy connectors that you are having. And now with the Raptor DB, you can have exponential improvements over performance dedicated to AI, actually. And ServiceNow is just keeping this layer as the foundation, for execution. Okay? Learning is, again, a loop that it relies on the data that you are having, the your CMDB, the relationships between your services, the definition of that. So this layer is key is fundamental, for the success of the autonomous workforce. And you have the whole the entire platform just covering, the the this capability. Well, I think most of you are, just waiting for the live walkthrough, so I will go to my instance here. Let me change here. K. And you are seeing my my screen now. This is, the service operation workspace. You can see here that we have, some view. We have already activated on this instance, our worker performance. So I have some data from the outcomes by category. So the amount of inquiries, help, or the amount of software, tickets that was reassigned or reopen it after, assigned to my AA agent, and some exchange resolved it by category also. But the key thing is if I want, for example, to check how efficient is so I am checking I am going for the outcomes. I'm going for the value that this agent is, adding to my enterprise. So the meantime to resolve incidents, for example, here, is actually, going down and compare it, with February. I am reducing 55% of the time to resolve incidents with my agent. The time, just call, go down with the first call resolution. Okay? Then we can go here for the value, and the value is just showing me here. For example, simple metrics that I have defined it. How is the sentiment of the users just with the tickets resolved by the agent? And this is improving here. The adoption coverage, this is the more data that we have, but it's a number that I can track, on, how it's the coverage increasing or not if I have some feedback. And then if we need, to add more specialists, I will zoom in here. You have here. This is the autonomous workforce, and this is specific. The role, again, is a level one service desk AI specialist, and I can activate it with a single click. The definition is here. So you can see that, this is a a team worker that is has some expertise, role based for the task that is, resolve, incidents or tickets here from the service desk. It's always available. Okay? Can participate in communication channel. I can, connect to workflows. Okay? And let me activate. So just to activate here, it's asking me if I agree the skills that this agent at the moment is available to resolve so I can, decide, okay, can resolve VPN issues, can resolve laptop issues, can work on these groups that I have on my instance. Okay? AI inquiry team, AI case team, level service desk level two, service desk. So and the roles that has permission. So I added some roles here just to confirm. It has the required permissions, and only with these permissions, I I know which is possible or not to execute. Okay? Just confirming this, I activate. It's really simple, and now it's active. Okay? So let me change here my screen, and let me move here to Moveworks. So, I don't know if you are aware, but, employee works came from Moveworks, actually, from the acquisition, and is the new from Dart for ServiceNow. So I'm just testing directly here on Moveworks. So let me type here find an answer here. So let me say I am unable to, access some more applications today, and I have an error message here saying it's not able to to connect. Okay? So the first thing, the assistance is just checking. Okay. Give me an answer here. Let me check this answer. So the assistance is just telling me, okay. I understand it according to the description, identified that it's involved VPN remote access. So this is the assistant part that we mentioned it in the beginning. So let's try this. Check my VPN. Let's try another this, and we have this the knowledge articles. Where is this coming from? If it doesn't work, you can do this. Okay. So this is an assistant. Could be helpful. It's not, just a virtual agent throwing me a entire knowledge article, but this tailored they answered to the answer to the things that it's actually related to my problem. But I will say troubleshooting step didn't work. Have been raising an incident. K. So working, raising an incident. Filling your ticket. Okay. Let me go through this. So I asked it to raise an incident. I understand you want to raise an incident. Okay. File ticket. Fill in my ticket with the details and created this incident. Okay? So I have one action here. This could be a workflow, but I have the action of, just fulfilling my incident. K. Thank you. Let me go now to the incident itself, and let's say that the autonomous workflows let me close this. Okay. Now we change it to the service operation workspace, and we have our incidents here. Beautiful. Let's raise it. And look. It's assigned to the AI level one service desk specialist. Okay? So it's not a human. It's actually defining and assign it to an AI specialist. Let me do through this. Okay. Now I'm seeing the incident here. I'm unable to access applications. We have the summary, and we have some actions, actually. I rise at the ticket, and look look at this. The configuration item was defined, and we have some, fields fulfill it. Okay? So if I can go here to details okay. Let me check. You can see that the description was fulfilled. The category was identified as subcategory, so the the agent just defined the category for this, identified and defined the configuration item, fulfill it by AI. It's clear here. Okay. It's assigned to an AI specialist, so it's not a human working on that. Okay. Cool. Let me check here. So I have AI workflows here, and it's checking remaining steps. Okay. So it's working on my incident. Let me check for the details here what is happening. Okay. I can see a lot of work being done here. So started an agent, fetching details, fetching the task, remaining steps, make the triage, here. So defining the solution and okay. It's just updated with some resolution notes. Let me check here. A agent just added some comments. Hello. And your ServiceNow AI, I'm helping your incident. And I have some resolution notes. And you can see the ticket is now resolved. L one service desk got modified. It's resolved. It was identified as moderate, and resolution was provided, and there are some resolution notes. So I'm your ServiceNow AI helping in the incident. The issue was identified. Okay? The solution was proposed. We have already taken the followed automated action in behalf. Syskala services was reconnected, with success. So it verified in if I check, executed the DX remediation, okay, and then started the to call the closure agent. So it actually takes some action, verify that the action was actually succeed, and then, you can see that it was a health check, and the report was actually, successful. And then I can have some evidence that the problem is gone. And then when the problem is gone, the incident is actually not anymore, required and was solved. Obviously, as a user, I can check. Okay. The solution is good. Now I can give some feedback, but it's actually resolved, this incident. Okay? So this is the power. If you have some questions, please, you can add here, and I can see, what is happening. The the thing is done. The the work is done and tested and validated using the defined guardrails, the defining security, definitions that we have. But using the workflows that are available for this agent to to use. K? Well, let me go back here. This was the the live walkthrough. Let me check just, time. Okay. We are good in time. Now, some details, how this level well service desk specialist work. You can define the rules, but this prepackaged it from ServiceNow and and Moveworks team together. Self assigned tickets. So once you go to this screen to activate, you can just add to the team and you define the teams. Yeah. We we saw this. Can identify, understand, resonate about which is required on the ticket and resolve issues end to end, connected to your CMDB, connected to your knowledge base, and could be escalated with the full context required. So you don't need exactly to redefine a new process or reinvent something. It's just adding more people, digital people, to your work. Okay? So this is why we are talking about workforce. Right? It's autonomous, and it's ready to jump in your process using your tools, using your data, that is existing even for the humans and is taking advantage seven, twenty four seven. So, all all week, all time is available to work. Then it's just a simple workflow. Is is supposed to follow this. So it's AI with predictable outcomes that you can rely on, analyze the request, checks if you have some known issues, review device history, prepare a diagnosis to try to identify the root cause of the problem, execute fixes directly on the employee device. For this, I just want to to comment, it depends of DX. If you are not aware of that, you can you can check the documentation, but you have some device execution requirements just to work on the device, but it's possible. Validates resolution, notifies the employee you saw that, added some comments and was the text generated, but is actually telling the employee what happened, what actions was taken. And we have, in the other hand, the full log, the tailored log, if you, as an admin of the platform, needs to go for this, and documents everything in the ticket. And, also, if it's required, the knowledge article, could be updated. For example, the solution is not clear, but maybe on the knowledge, article, you have a knowledge gap or some suggestion just for someone to review, check, and validate, and publish an updated knowledge article, but it actually updates the knowledge base. And as you saw, resolve the incident is just some few minutes. Okay. Road map. What's next? You saw the level one services specialist was, actually, launched in April. ServiceNow had an event. If you miss it, you guys you go to servicenow.com page. You can see again this available. And during knowledge and the the the rest part of the year, the second half, you will see another workforce autonomous workforce coming live. So we can expect an specialist to take a task and work for AI ops world, AI operations, and IT operations. You will have autonomous workforce taking, the work for the software asset management. So it's an specialist that makes some part of the asset management work done. You can expect a security analyst, and you can expect demand management analyst in the world of project management and helping your PMO, just getting more work done. Okay? And this is only, part of the, the initial, roadmap that we can see. Just sharing. If you go to ServiceNow, I actually has the the the the site, servicenow.com, autonomous workforce. You can see live, data. So value of autonomous workforce, how ServiceNow is telling. And you can see here the specialist that you can, receive every corner of your business. So the IT case for CRM, what is expected for employee experience, level one, HR service delivery, AI specialist. Right? Risk and security. And for app development, you already have this one. This is a best friend for all of us developing applications, the build agent. So it's autonomous creating applications. But every corner of the business has now agent a agentic workforce. Actually, autonomous workforce. Right? Well, we have some minutes, still. So ready for questions, answers. Yeah. Hoping you have enjoyed, today. Let me just, finish here my sharing. Amazing. Thank you so much, Diego. We did have a couple of questions come through, so I'll just go through them one by one. Can we use agents to predict field based on existing record instead of predictive intelligence? Okay. Let let me think a bit. Can can you repeat, please? Can we use agents to predict field based on existing record instead of predictive intelligence? Okay. When you have predictive intelligence capacity, or the capability for predictive intelligence, it's a framework when you have to use machine learning for a set of records. You have different structures or methods to use predictive intelligence and fulfill, some fields. Okay? This can be it's not exactly replacing, but you can use on the autonomous, workforce. The level one service desk agent can take advantage of the data that is predicted by predictive intelligence, could use that, but, actually, could be more context defined. So you can expect different behaviors for each one of the technologies, and I can understand that the autonomous worker has more context and live context, than the predictive intelligence because it depends of the data that you are just giving to the predictive intelligence model, that normally it it start, for example, the last 10,000 records is and this needs to be executed. So I think it's different, just that. Okay? Amazing. Thank you. Is employee works and auto the same? No. Actually, this is a very good news, from knowledge. ServiceNow launched at auto. OTO is the AI experience. Okay? And which means AI experience. You have in the past, like, almost ten years ago, virtual agent launched. Okay? It's a chatbot. Just that. Then you have Now Assist that is an AI assistant as and and we talk about AI assistance here. Anoto is not only chatbot or texting, experience. It's actually AI experience because can see can use see your screen, can, use voice. So we have voice agents and can chat and and talk with with text. So auto is actually the experience layer. Okay? And the autonomous workforce and the this level one services agent is one part is not actually changing the experience, but is the agent that came to execute work, to understand the context and execute work, define it. And we can have, given the road map, different type of workforce. But auto auto just enter it to be the new experience for all employees or all workflows even if you have or you have not activated any, autonomous, employee or worker. Okay? Lovely. Thank you. Does the l one service desk agent come pretrained? Yes. ServiceNow worked a lot, and movers team worked a lot also. So have a very good level of knowledge, on how to solve common issues, but it also relies on your data your process, your knowledge base to learn. So you have some AI index, and you have past incidents from your company that the agent can use. So it's a specific context, but, yes, indeed, have some pretraining. Just it's not started from day one and, like, an employee that, this is my first day working and never solved an incident, could be theft. Now the the agent is just clicking, adding to your team, and has of a pretraining knowledge to resolve incidents and, yeah, Lovely. Thank you. This question is referring to the demo portion, tasks. and they wrote, how the category and sub subcategory filling, is this based on a prediction or ML? Okay. It's actually a prediction. If you don't have the machine learning part, before, it's it's actually on prediction. Understand, again, context is king. The context of the question, the context of the user, the context of the type of, incidents, and the categories that you have on your instance, you have a prediction score decision, could be wrong, and you can train. Okay. This is with feedback. This is not the expected, category, but it depends on on our prediction and the highest level of, according to the context. Lovely. And then I think this is the last one. Does that need a special agent on devices to permit AI agent to execute actions? Well, first, it runs in the platform itself. So if you are using, or try to solve something, you just ask think on that. Let me rephrase. The service desk manager or the service operations manager just needs to add a new workforce to your team. So you have, I don't know, 20 people working, humans, and you are adding one, two, three new workers, digital workers, and they will just, executing the work. So it's not depending on the device you are executing. Now the other part is to solve something, I need to execute something in the, for example, in the desktop, laptop, or the mobile from the user. Right? So if you have the workflows integration required to execute this, it is done will be executed. Okay? Lovely. Thank you. And, actually, another question came through. Mhmm. If you have knowledge based data, but it's not AI indexed, what is a smart and OOTB way to prepare it for auto slash AI usage? Okay. Yeah. Actually, AI is part of the platform, so I can't imagine I have some knowledge article in the it's not indexed because you have the engine out index. Otherwise, you have, for example, this is secure database, or or knowledge base that should not be ingested or should not be used if you have some guard guardrails for that. Obviously, the the autonomous workforce, is is the the level one services agent won't take that that data. Right? So you need to work on governance and security definitions just to make this, working. But, yeah, if by default, if you create something and it's available for the agent to consume, will be indexed. Lovely. Thank you. I think that's all the questions that we had, so I'm happy to wrap up the session. Thank you so much, Diego, for your time and for sharing, all of your knowledge. We've got some nice comments coming through as well. So, yeah, I'm pleased that it resonated with everyone in the audience as well. Thank you for spending, the past hour or so with us and for sharing all of your questions. As I mentioned in my introduction, the recording will be shared in our follow-up email, so you can expect that in the next couple of days. And, yeah, hopefully see you in one of our next webinars soon. Thanks, everyone. Bye. Thanks so much. Bye.