Hey, welcome everyone. I'm excited to have Amit Zavery, president of ServiceNow, join us for this session today. We'll talk about the future of AI, AI in the enterprise, AI for developer productivity, the future of workflows — a lot of interesting topics here. So Amit, before we start, how about you quickly introduce yourself to the audience here.
Joi, thanks for having me. I'm very excited to have a chance to have a conversation with you. So I'm Amit Zavery, president, CPO and COO of ServiceNow. I run product and engineering, and the strategy and operations of the company. So quite actively involved with all the things we're building, and the partnerships and the discussions we've been having in the industry as well.
Well, great to have you here chatting with us. So let's start with AI in the enterprise. There's of course a lot of conversation here. We sit in Silicon Valley and there is not a single conversation you can have without AI. But what are you seeing in real enterprises? ServiceNow works with thousands and thousands of companies. What's adoption of AI? What are the key use cases you're seeing people adopting?
No, you're right. I think AI is top of mind for everybody we speak to, right? So every customer I go and speak to, the first thing they want to talk about is what to do around AI. So the biggest question for many companies right now is where to start, what to do, how to get it done, and how do you create value out of it. That's kind of the typical conversation. And good thing we've been fortunate at ServiceNow to start in the AI journey many years ago with the Now Assist product, which has been around for a few years. It was the idea of helping customers as they adopt workflows and integrate different parts of businesses, and they want to automate business processes. AI became a very integral part of that. So ServiceNow has been delivering AI capabilities in that mode, and over the last year, year and a half, we've been kind of accelerating what we do around agentic. So agentic has become top of mind for every customer you speak to, because enterprise customers want to do automation. They want predictable outcomes, but they also want the ability to really improve how they operate as well as get value of the investment they're making. So agentic really clicks with them, because it really is giving them the ability to understand the business processes, automate those business processes, figure out where to make changes, and automatically make that happen in the back end.
Yeah. And is the primary driver efficiency for them? When you're talking to these customers, they want more efficiency.
Yeah, efficiency is kind of the starting point, but I think efficiency leads to eventually flexibility as well.
So they want the efficiency because they want to kind of the arcane systems they every legacy every enterprise company have a lot of legacy systems, right? And for them to kind of take those legacy systems, modernize them and have to do that in a much better way than they used to do typically, AI might be the kind of the catalyst for them to go get going.
So the efficiency of taking that existing investment, improving and modernizing is one part of it. The modernization becomes a part of the conversation, and second is to really have a much more manageable environment which gives them the way to operate differently, improve their bottom line for sure, but also improve the top line by customer satisfaction, other things they want to also drive with AI.
Yeah. So you know, of course a lot of people in the audience are developers and looking at software engineering problems through AI. What do you think of the adoption of AI for the software engineering workflows? Of course, AI for coding is everywhere and everyone is using it. How are you guys doing it internally from software engineering? What are you hearing from customers?
Well, for sure, I think as you mentioned, we are every company's thinking about how they can improve their operations, right? So we have something we call Now on Now, which is like taking a lot of the learnings we've had with the products we built and bringing it internally inside as a kind of making sure we're the client zero of adopting a lot of the technologies we build and sell to customers. Second, we of course work with a lot of partners, including you all, about how do we improve the efficiency of our operations for engineering as well. So a lot of people talk about vibe coding. There's a lot of opportunity to really improve efficiency from how you build and test and deliver software. As well as what we're doing now for customer service becomes very important. So we have done today around 76% deflection rate on customer service. So we have a large case volume go up because our business has gone up very fast. But our customer support team is not growing anymore. We don't need to because we've been able to handle a lot of things with customer automation and self-service kind of things. Similarly, we're doing that for sales operations, finance. So every part of the business is having some impact because of the AI efficiency we can bring in. And the software engineers today, we use coding tools. We of course now releasing software much faster because of the efficiency we're bringing with AI capabilities. Now a lot of ways to test better, which is important. A lot of engineers, as you know, don't want to write a lot of documentation, so AI helps them with that, but also creating unit tests and things like that as well, reuse of code, as well as integration with various dependencies and things like that. So it's impacting every part of our life cycle and every part we touch internally between different groups as well as everything we do with customers and our partners as well. So AI has become kind of the common theme and there's not a single area which we're not rethinking.
Totally. You know what, that's what I'm starting to believe. Like, you know, if someone doesn't believe in leveraging AI in their jobs, they're probably doing a disservice to how they work and how they operate because the productivity gain is real, the process impact is real in many, many things. You guys talk to a lot of customers. I'm sure you have seen this regularly.
Yeah, we definitely like, you know, we see like the first step of AI in software engineering, AI for coding, which is very obvious. Like, you know, the coding impact of LLMs, you know, everyone is using in some ways. But we also see an interesting dynamic. Like, you know, the coding by itself doesn't really increase your throughput, you know, of how fast and how frequently you're shipping. Actually, the folks at Google, they have the DORA research team. The DORA research team published this data this April in the Google Next conference, where they modeled everything on like teams using AI for coding. They said it's a 25% increase in the code volume, 25% more code coming out, but the throughput decreased 1.5%, and the stability decreased, I think, 7.5%. So if you have AI for coding, the code will have more bugs and vulnerabilities, not because AI is inaccurate or not, because no developer can read all that code. There's too much code coming out, and if your downstream processes of testing and deploying it and security and governance and all are not robust enough, it's not going to get more throughput out of it.
Yeah, you're right. I mean, software engineers are not just writing code. Yeah, right. They have a huge amount of other parts of the task. So you have to automate all of the things. We've been starting to look at every part of their life cycle and making sure that we can provide tools to software engineers so they're more productive. But we've also been thoughtful about what you can do as well, right? Because a lot of the software engineers, they have their own ways of working. You can't just force everything into that kind of conversation. So things are changing, but I think the AI technology definitely is working. It's more of now how do you adopt it. There's a lot of cultural changes required, of course. It's not just a software change, right?
Definitely, it's a cultural change. In many cases, it might be just a job role change. Like, the job role and the job discipline might be totally different. You're right. I mean, we today, for example, support all the large language models. Last year, we started with our own domain-specific models, and now we have all the frontier models out there in the same code base, right? So we're kind of making sure that all these things can be tested properly with prompt engineering kind of roles. We're doing a lot of work in terms of how customers can select those things, so what kind of permutation combination and the whole configuration requirements are there. So a lot of changes need to happen there.
Yeah. Yeah. So you know, let me ask you questions around how people build enterprise applications on top of a platform like ServiceNow. Like the great thing about a platform like ServiceNow, and I've seen the evolution from doing certain things to now a full workflow platform. You know, you have Satya Nadella from Microsoft saying the enterprise SaaS won't exist because companies can just build their own solutions with AI. How do you guys think about that?
Good conversation to have, right? I mean, I think the software is changing, the technology has changed a lot. You think of ServiceNow as a platform company. I mean, that's how it has originated. It's always been that, right? This idea of one platform with one data model, one user experience, allowing you to connect every part of the business. So it's a workflow capability around at a business process level. So it's a really enterprise productivity kind of mindset versus a data or a particular application only, right? So the platform allows you to, one, build your own workflows on it. So we provide all the constructs and capabilities where you don't have to do all the heavy lifting. You can build your business process. So can we provide packaged business processes, workflows, right? Could it be for IT, for security, for HR, finance, customer service, sales operations, CPQ, those kind of things, right? So it's really a platform which kind of extends in that regard. Now it has the same platform as AI. So it's easy to now take AI capabilities. We have the whole agentic OS built in, right? Built around AI agents, the whole orchestration engine, the control tower for managing security of AI capability, but also the trust and safety and compliance, which is very important for enterprises, right? So the idea that we can provide you all of this stuff in the platform, and then you can build your workflow or integrate various systems together with any kind of business backend as well as any kind of data platform. So we connect various data systems together, but also any kind of application workflows, right? So the difference for us versus maybe this thin veneer of SaaS is very stark, right? When Satya talks about SaaS being replaced, a lot of the shallow capabilities is doable with large language models. But once you start doing these enterprise business processes, which has a lot of context, which has a lot of security requirements, a lot of requirements around compliance, trust, safety, you know, those financial institutions, regulated industries, or even any kind of manufacturing companies, or you talk about any software companies, these processes are very complex, and it requires a lot of depth of understanding as well as you have to change your businesses quite often, right? So a lot of the domain of understanding is not just in the business context in the application, it's in humans as well. So there's a huge amount of interaction required with humans, but you can't codify everything. And then the whole things you require in a platform. So that's what we solve for. So sure, there will be pieces which can be replaced in shallow SaaS applications. We don't think of ourselves as a SaaS application because we're a platform company. You can build your own, of course.
Yeah, totally. I 100% agree with that. Like, you know, I feel like it's a bit premature to have that. Maybe it was a bit... I think it's always exciting to have those debates. Yes. Right. Nothing wrong with it. I think people get a chance to think about what the future will look like, but I think there's a reality in terms of what enterprise software is, right? And you and I have been in enterprise software for ages. And when you look at these enterprise companies, there's a large amount of investment which they made, and these systems are not that straightforward to go and replace with a single line, a few lines of code, right? There's a lot of... Also, enterprises need, like, you know, deterministic nature and also need auditability and controls and governance. Like, when we look at the areas that Harness, we have, you know, really good AI for, you know, creating deployments, like your CI/CD pipelines, for example, right? So many times people think, like, is AI going to do the deployment in production for me? And we tell them, no, AI is not going to do the deployment. AI will create the deployment pipeline that's auditable and governable and meets your enterprise standards. But all the burden of creating and maintaining that pipeline, AI can do. But it's the pipeline which will do the deterministic deployment for you, so that you have all the enterprise controls and governance, and so you don't have to fear it. But that layer is still needed. Like, that is the fundamental layer that will do that.
