GenAI in the Enterprise: A Conversation with ServiceNow

Dave Nicholson:
Welcome back to Six Five Summit. Dave Nicholson here. Got a very special guest, Mr. Jeremy Barnes. He’s vice president of AI platforms at ServiceNow. Welcome the program, Jeremy. How are you?

Jeremy Barnes:
I’m great, thanks, Dave. It’s really great to be here.

Dave Nicholson:
AI, something that we’ve heard a fair amount about over the last year or so. If you accept that we’re at least a year into this gen AI phenomenon, how would you characterize where we are in this journey?

Jeremy Barnes:
Well, we can say that we’re a year in from gen AI, but really AI has been going on since a long, long, long time before that. Sometimes people say, “What’s new about gen AI?” I’ll say, “Well, it’s kind of like AI 2.0 in that it’s AI, but it actually does what you expected it to in the first place.” I think we’re in a very exciting part now because before, AI was something which people kind of put in the dark corner and wanted to behave itself and not spend too much effort on it. Now it’s in the mainstream and people are thinking, “AI is able to do things that I’d only dreamed of a few years ago.” Overall, we’re in a pretty good place. It’s pretty fun. It’s exciting to be part of this.

Dave Nicholson:
So specifically gen AI as a subset. Of course machine learning has been going on for a long time. Now you’ve been at ServiceNow for five years, is that right, somewhere in there? Is that right?

Jeremy Barnes:
I’ve been here three years. I was the CTO at Element AI, which was acquired about three years ago. Yeah, three years, been a while.

Dave Nicholson:
Okay. Three years, but in the recent past you’ve talked to a fair number of CEOs in this space. Is that fair to say, that you make the rounds?

Jeremy Barnes:
Yeah, absolutely. CEOs, but also there’s a lot of subject matter experts where a lot of CEOs know that they’ve got to do something but don’t necessarily want to take the time to learn, and so I often get involved here with CTOs, CIOs, and their teams in order to go one level of detail down beyond, “What is this AI thing?” To, “Exactly what can we do with it in our business?”

Dave Nicholson:
What are you hearing as the kind of hopes, dreams, aspirations, and concerns out there as you engage with folks on the front lines?

Jeremy Barnes:
Well, certainly let’s start with concerns. Absolutely the mandate in this year… 2024 is a year of board mandates to go figure out generative AI and what it means for the company, and so a big part of the concern is there are results becoming expected. This is not just to do a proof of concept and make a nice graph. People are thinking, “Okay, what is this going to do to my business? I have some challenges with my business that this might help me address. I want to actually move the needle on those things. I don’t want to just sit down and talk about it.” So definitely concerns about, “Are we going to be too late? Are we going to lose competitiveness if we don’t move fast enough? Are we going to move too fast and break things?” There’s a lot of that.

Hopes and dreams, I mean, that’s the part which I find the most exciting, and that’s really… There’s a lot of… There are kind of… I call them snowplow roadmap. It’s like you imagine a snowplow driving and the snow’s just getting pushed in front of it. There are a lot of things in companies which were, “This is part of our vision. I think someday we’ll be able to do it, but we keep just pushing it off a year in front and it just stays in front because it never gets away from the snowplow.” Now gen AI allows us to clear that, it allows us to clear that jam, it allows us to do things which were potentially years and years out before, so we’re also seeing a lot of reconfiguring, which was, “We had structured our organization…” Here’s what I’m hearing, “We structured our organization based on some things being very hard and we think that now they’re going to become really easy and suddenly there’s these gigantic new opportunities which we’ve got to take, but we didn’t set things up for them in the first place.”

So all this change management in order to allow these hopes and dreams to be realized, we’re seeing a lot of that as well, so the concept of doing a digital transformation in order to drive something. A lot of people started off on the digital transformation route with a sense of, “It’s going to help with someplace on the bottom line, it’s going to make our workforce more effective, better serve customers.” Gen AI is clearly the killer app for a digitally transformed enterprise, and so now it’s like, “Okay, we see that. We got to just put this in overdrive and get everything ready from the technology perspective, but also the organizational perspective as well to make this a reality.”

Dave Nicholson:
Yeah. Now, it’s interesting because ServiceNow has been a trusted partner for businesses for many, many years, and in a sense ServiceNow is on the very same journey that its customers are on in terms of coming to grips with and leveraging generative AI moving forward. What’s the current state-of-the-art in terms of how ServiceNow is infusing what it’s been doing for its customers for a long time? How are you infusing it with generative AI moving forward?

Jeremy Barnes:
Yes. That’s a really great lead in because this is something which we’ve put a lot of effort into building out our CTIO department. We do two things. Number one, that department there is empowered to do the same things that our customers would do, so not necessarily using the tools that we build but looking across the broad industry ecosystem and saying, “What is the best tool for the job? How can we make the organization as effective as possible?” But at the same time, in parallel we also deploy all of our products internally and we are customer zero for them, so a lot of the early feedback, validation, testing, a lot of our numbers are in terms of, “What does this mean for a business? What can I expect in terms of outcomes?” Are calibrated based on what we see internally as well-.

Dave Nicholson:
So you drink your own champagne, as they say.

Jeremy Barnes:
We absolutely drink our own champagne. Drinking champagne too early doesn’t necessarily taste as good as when it’s been aged for a while, but that gives us the opportunity to know… We absolutely don’t want our customers to drink it until we know that it’s a really good vintage, and so that’s the approach that we take. We’ve put a large amount of effort into building out this CTIO department, which can be completely empathetic to customers because it’s got access to the same tools as well but also understands exactly, “What is the gap between what’s needed and our products or what’s there in the industry as well?” That’s a really big advantage that we have in the ServiceNow product development methodology.

Dave Nicholson:
In general terms, when people think about information being generated in the generative AI context, which is different than machine learning coming up with a response or business guidance, often people are concerned about this idea of bias entering a system. Not necessarily political bias, but any kind of bias that can direct guidance in a way that we don’t necessarily want it directed. Do you see bias as a challenge? What about governance and things like that? What are some of the challenges around generative AI from a ServiceNow perspective?

Jeremy Barnes:
Yeah. We absolutely care about those things for the simple reason that our customers do. One of the ways that we think about product is it’s not just a question of what does a product do, it’s how effectively can it be used. For that reason, we think about those things so our customers have good answers to the questions when they’re asked them. For bias, I think the thing which a lot of people don’t understand about it is if you don’t… Everything has a bias, so if you don’t measure it and design so that that bias is minimized or so that bias is at exactly the level you want it to be, there’s a bias there, you just don’t know where it is.

A big part of it is how do you do that testing? How do you proactively look to see, “Let’s assume that this is biased. What could it do?” Then test to see if any of those outcomes are happening and to the point where you can say, “Well, we are now pretty sure that we know exactly what that level of bias is and that’s at the acceptable level.” We have hundreds and hundreds of people who do testing of all different kinds to ensure that we’re able to make that promise to our customers, and that means our customers don’t need to have those hundreds and hundreds of people in every single one of our customers there. Bias is… It’s a really interesting and subtle subject. We don’t expect our customers to be experts in that. We take that work out and we… That’s part of what we provide.

Governance and everything around responsibility and trustworthiness of AI, there’s a few things that we do there. The first is that we have our human-centric AI guidelines and so things like transparency, human-centricity, accountability, things like that. They ensure that when we develop our products, that we look at them from all the lenses which allow us to understand how people are about to see the products and how we… Trust is not something that is automatic, so trust is something that you build up over time. Making sure that we allow that trust to build and we think about exactly how that’s going to happen.

Governance, you’ve heard, I’m sure, about all of your EU AI act for European customers, there’s the presidential order which is beginning to come into effect, some of those rules have been published, and then there’s just all the normal enterprise stuff you do around risk and governance and things like that. Our customers… We have very large, sophisticated customers.

They absolutely care about those things, and so it’s also building into the product experience the ability to understand what’s going on, to surface the information, and to govern those in the existing enterprise processes, which allow it to be deployed in a way which is compatible with whatever procedures and policies that our customers have in place. That’s all the building around it. That’s not the core technology. That’s everything you need to build around it in order to enable effective and successful customer deployments.

Dave Nicholson:
Yeah. Yeah, that makes sense. Well, let’s talk about more kind of… I don’t know if it’s necessarily the core, but just can you give me some examples of… Maybe look at it this way. Something from the ServiceNow portfolio or suite of capabilities, what that scenario traditionally looked like before AI. Now we have the advent of maybe more advanced machine learning, now we’re talking generative AI. How do these things change? If you have specific customer examples, great, if you’ve got anonymized examples, great, but give us a more concrete example of what people are doing with generative AI in this context we’re all familiar with, or most of us are now, playing around with the various tools that keep coming out every week. But what about if I’m in an IT service management environment, what’s generative AI going to do for me? Can you make it more palpable?

Jeremy Barnes:
Yeah, absolutely. I really love that question because we are fortunate enough that there are millions and millions of people who spend all their day doing work and getting work done inside the ServiceNow platform, so when we talk about how do we make this concrete we can actually talk about specific user personas that spend their whole day using the software that we build and we understand what they spend their time doing, where they get stuck, when they meet their objectives and when they don’t. A lot of our prioritization that we do in our roadmaps and things like that is related to not what we imagine that people want to do, but really the work that they’re actually doing and how they get that done. An example here, I’m not going to go into the obvious virtual assistant type example because I think that’s been done to death a little bit, although we have a really good story there. Let’s talk about-

Dave Nicholson:
Yeah. Give us something else.

Jeremy Barnes:
Yeah. Yeah. Let’s talk about a service agent. Your responsibility is to… You’ll have a set of incoming incidents or cases that you need to handle. This could be an IT service agent, this could be an IT operator who’s looking after a data center, this could be customer service, and there’s a whole bunch of you know, services, so there’s a lot of personas who fit into that.

Now, if you look at how people are working, one of the biggest challenges that they have is if they just finished their day and they’ve handed over to the night shift or something like that, they come back, all the things they’re working on could have changed throughout the night or there could be new ones that have been handed over. There’s all these reasons that suddenly their idea of what’s going on might not be up-to-date with reality.

And so, in the interface, we allow them to… Various productivity tools there coming out, things that allow them to, for example, figure out what they should work on next. Once they get it, it’s a summary of what happened since our last one so that instead of needing to go back and sometimes there’s hundreds and hundreds of interactions, they get a sense of, “Okay, these are the important things that have happened.” Once it comes time to resolve it, there’s probably information elsewhere in other incidents or cases which are pertinent, but they’re not going to read through 120 of them to understand exactly what’s going on. The generative AI can do that and create a summary saying, “This is what’s common, and in your situation this is what seems to be what’s happened,” and present that information to them.

Once they get to the point where they want to interact with the user again, instead of needing to type out all of the details they can be taken from that summary and put into place. Of course they check the message before to make sure that it makes sense, but it’s a lot less mental effort to do that, and ensure that they didn’t miss anything as well. They send that and let’s say that the incident is resolved after that. Their incident notes can be created with generative AI in order to make sure that the resolution notes are accurate, but that that information is available for the next agent as well. Then finally, if they realize, “Actually, this is something which is happening frequently,” and they want to create a knowledge base article, they can create that using generative AI and that can feed directly into auto-resolving incidents. When a user will come in through the virtual assistant, it will be able to use that information to allow them to self-service afterwards so they can also create this extra efficiency there with the automation.

Now, all of those things you could have potentially done as a service agent, but the time it would’ve taken to do them all means you never would. You cut corners. Productivity’s important. We enable a faster outcome, less mental load on the people as well, and each time that an agent interacts with the system they create a basis for further efficiency in the system itself. It’s really generative AI applied not as just one use case, but in lots and lots of places that enables all of that to happen. We find that really, really exciting because we can look after really the entire workflow of people in our platform.

Dave Nicholson:
Yeah. That’s a great example of having the digital assistant that’s there to basically say, “Hey, welcome in. Let me tell you what’s been going on since you were… Let me get you caught up so you can hit the ground running as you get going.” I know personally I’ve started using very, very informal language and tone of voice when I’m interacting with these natural language assistants. It’s amazing. Well, tell us about what is… What do you see coming in the future? We’ve got about a minute left, if you can kind of give us a summary of what you think the future holds.

Jeremy Barnes:
The future is really interesting because it’s not going to be a future of just generative AI. It’s not going to be a future of people doing one thing and generative AI another. It’s really going to be a future where generative AI, as agents or as assistants, works together with people and that’s going to push the horizon of what people expect to be able to get done, the kinds of problems they can solve, the things they can build, the level of service they can provide. It’s going to increase the horizons of people’s ambition, so that’s going to be really exciting. All of that is going to happen.

We also see it from an enterprise perspective. Like I said, this is the killer app for a digitally transformed enterprise, so enterprises are going to put the digital transformations into overdrive. They’re going to use that to do things with generative AI that make the biggest step change in their business that they’ve ever seen, and we’re going to see that driven by this technology but by also this maturity as people understand that generative AI is not something you bolt on, that AI is something that you build into your business. We’re going to see this. The businesses that do that well, we’re going to see them just achieve things which are absolutely amazing, even by today’s standards.

Dave Nicholson:
Sounds good. I say we schedule a meeting a year from now. Let’s have our digital assistants meet one another and discuss what’s happened in the prior year, and then they can brief us individually. It’s a brave new world we’re entering, my friend. Thanks.

Jeremy Barnes:
Yeah. Our avatars can chat for sure about that.

Dave Nicholson:
Thanks so much, Jeremy Barnes, vice president of AI platforms at ServiceNow. For the rest of you, stay tuned for more interesting content from Six Five Summit.