'You need the right data to get the outcomes you want, AI is not magic, it's maths.' – Dave Wright, ServiceNow

Tahawul Tech

19-05-2025

CNME Editor Mark Forker spoke to Dave Wright, Chief Innovation Officer at ServiceNow, during Knowledge 25 in Las Vegas, to learn more about how the global ITSM leader is going to use their platform to power the new era of Agentic AI across the enterprise space, why they are viewing agents as 'assets', the importance of cleansing your data – and why his innovation office is focused on futurism.
Dave Wright has been a central figure in the seismic success of ServiceNow, who's stock as a company has rocketed both in a figurative and literal sense over the last number of years.
In this new transformative era of Agentic AI, ServiceNow has emerged as the company with the platform to power this movement, with NVIDIA CEO Jensen Huang describing the company as the Central Nervous System of the 21st century in the enterprise space.
A big announcement on Day 1 of Knowledge 25 was the launch of ServiceNow's AI Control Tower, and Wright outlined how they are viewing agents as 'assets'.
'We now have the capability of building agents that actually do something that can affect something on the background, but there's got to be a way to control that. However, the challenge at the moment is the fact that nobody has built agent detection technology. It's not like you can scan an enterprise in the same way you can scan networks, so what we wanted to do was build a Control Tower that will allow people to not just manage ServiceNow agents, but manage other agents, whether it be on Copilot, or Salesforce, but we were asking ourselves how are we going to do this because we can't detect them. We actually use the concept of putting them in the CMDB, and we treat the agent as an asset. Essentially, what that means is any agent that's written on any platform is seen as an asset and what allows us to do better understand the relationship between the assets. We'll see the links that exists between agents and similar to how you can manage the lifecycle of an asset, then you can manage the lifecycle of an agent,' said Wright.
Agentic AI is only as effective as the data you feed it, and in 2024, 60% of AI projects were abandoned due to incomplete data, so the governance around the data is going to be a tough nut to crack, and Wright has advocated for companies to have a Chief Data Officer.
'I think if your going AI-first, and that's going to be your strategy long-term then I do think you need a Chief Data Officer, but it's always hard to say that implicitly because you'll always find someone who has done it without a Chief Data Officer. It also comes down to the size of the organisation too, if you have a huge company with a lot of data then you're going to need to be more cognizant about all of that. The big challenge is the fact that people have different perceptions about data. I used to get people come up to me all the time and say our system is going to be amazing because we've got 10 years worth of data, and I'd go that's great, but has anything changed in the last 10 years, and they'd go yeah loads of stuff has changed, and then I'm like well that's sort of useless. Because take for example, let's say you have a process in place for opening a new job rank, and you've got 10 years of data and you have changed that process three months ago, then 97% of the time it's going to recommend the old process because that's where all the data is. The interesting thing about AI is if you said how do I open a new job rank then most likely it'll give you the old way of doing it, but if you change the question and asked what's the most recent way of opening a job rank, it'll give you the right way, so you need to provide the right prompt,' said Wright.
Wright highlighted that another major challenge businesses have is cleansing their data in order to ensure that they are using the right data when building their datasets.
'When we first deployed artificial intelligence from a machine learning perspective, I had a customer come up to me and say we deployed your predictive capability categorisation and it is not good. He said that 51% of the time it's right, but 49% of the time it is wrong. We then looked at the data they were using to build the dataset, and the data was wrong 49% of the time. You need the right data to get the outcomes you want, AI is not magic, it's maths. You need to feed it the right things in order to get the outcomes that you want. It's similar to companies who say they are going to implement AI to filter resumes when they come into the company to make sure we get the right keywords, so you feed and input the data from your whole company, and all the AI does is determine that 45% of the company is comprised of middle-aged white men, so it focuses on middle-aged white men. Businesses have to determine the outcome they want and establish the ideal profile of what they are looking for and make sure they are using that type of data. There are five, or six different dimensions when you're thinking about data, but at the end of the day, it's all about the quality of that data, and you need to make sure you're using the right data in your datasets,' said Wright.
The conversation then pivoted towards cybersecurity.
Cisco's Jeetu Patel recently said that AI is the biggest challenge the security industry has ever faced, and Wright believes that the industry is lagging behind.
'I do think the industry is playing catch-up when it comes to cybersecurity. However, I think there's got to be a point in the future where people need to have the capabilities to not only identify agents, but identify what those agents are doing. At the minute, you could identify an agent, but to actually know what that agent is doing then the industry needs to define some kind of protocol. We are getting agent-to-agent connectivity on one level, but it's almost like thinking of it as an IP stack, so there needs to be connectivity on all these different levels, so not just can I invoke you and get a response, but actually find out what are you doing, what's your security protocol, where are you allowed access? I don't think anyone has come up with it yet, but it's almost like a missed standard that has to be applied to AI, and there has to be some form of security standard when it comes to AI. A lot of companies are looking at it and thinking is this something that needs to be solved locally within their domain, or is it something that needs to be solved at an industry level, and I believe it needs to be solved as an industry standard,' said Wright.
Wright concluded the conversation by highlighting the work he does in his role as the Head of the ServiceNow Innovation Office.
'Innovation for me, it's not a person, in fact it's not even a team, but historically, it was. I've been at the company for 14 years, so when I joined there was only 300 people, so you were doing multiple jobs. There was a point where I was managing mergers and acquisitions, product management and pricing to name just a few roles, but as the company got bigger we were able to break those roles up. What the real function of this team now is it's closer to futurism than innovation. My job is almost like an influence role, I can talk to people about what I'm seeing in the industry, but I can't force them to do anything. It's a really good job externally, and a really bad job internally. I get to look at all the cool stuff and then I have to convince people to do something with it. At the minute, the product side of the company looks out at around two to two-and-a-half years, but our remit is to try and look out for three to five years. What do we think the future of the industry is going to look like, what do we think the future of technology is going to be, and so on and so forth. Essentially, that's where we get into all these conversations about does agentic AI move us to a post SaaS world, how do we manage agents, do we treat them as employees, or do we treat them as machines? We do a lot of research around this, and at the end of this month we are going to start releasing papers that have our views on where we think the future of technology is heading. We have been so busy as a company we have never really had a futurist function, but we do now, and we're excited about getting our research out there at the end of this month,' said Wright.