LAS VEGAS -- IT ops teams from large enterprises were among the first in their organizations to use AI copilots and plan to spearhead the adoption of AI agents, according to presentations at ServiceNow Knowledge 2025 this week.
It wasn't long ago that IT ops teams were awash in hype about AIOps, including the "NoOps" concept that suggested they might be about to automate themselves out of a job. That never materialized for most companies, but between the lessons learned with AIOps and the accompanying trends of tool consolidation and platform engineering, those teams now have a strong foundation to take on generative AI, said Carlos Casanova, an analyst at Forrester Research.
"It's not just AIOps -- it's going through ITIL, change management, so they have the concept of process and flow," Casanova said. "Then they brought in ITSM [IT service management] tools and began to make things more efficient with automating notifications and routing requests, and a service map was already there."
Within IT ops, Casanova said that past efforts at hands-off automation could especially benefit ITSM and help desk teams.
"The reason you've seen it more on the incident management side is that it's the most visible: 'Oh my God, the system went down!'" he said. "It's also an area of the IT organization that is most comfortable with an analytics platform that routes information, circulates work, uses queues, addresses the alert fatigue issue, so they're probably the furthest along."
Platform team builds runway for AI copilots
Fannie Mae is one example of an enterprise where platform engineering laid the groundwork for AI copilots. According to Raghu Bellary, vice president of enterprise workflow technology at Fannie Mae, the mortgage lender's decade-long work to consolidate on a centralized platform and automate IT governance also began with ITSM.
"From there, we went to IT operations management transformation, which was implementing service discovery, business service mapping and a very robust CMDB [configuration management database]," Bellary said.
Raghu Bellary, left, and Tatyana Kelly from Fannie Mae present at ServiceNow Knowledge 2025, discussing how their platform team automated AI governance.
After that came a large-scale cloud migration for the company of approximately 8,200 employees, followed by a move to Agile software development and business process automation. Along the way, the platform team retired 49 legacy apps and installed 52 new ones, centralized on ServiceNow's Now platform, which also went through 15 version upgrades during the cloud and platform migration process, Bellary said.
"So, a couple of years ago, when GenAI [generative AI] exploded around us, the GenAI governance team needed to handle a very large number of cases but didn't want to inhibit the whole process," Bellary said. "They couldn't manage it."
The GenAI governance team approached Bellary's team for help automating requests and approvals for generative AI tools. The platform team did this using tools and workflows it had already developed, including a self-service portal, centralized corporate data repository, automated validation and approval processes and notifications.
"Using things like Activity Streams brings a lot of transparency to exactly where a request is in the approval process," said Tatyana Kelly, director of ServiceNow technology and business application development at Fannie Mae, during the presentation. "It's all done inside the tool -- there's no need to make phone calls, send emails or chat messages. It made the process more efficient for everybody in the company."
All of that lays the groundwork for bringing in AI agents, which will follow the same governance process, Kelly said in an interview after the session presentation. The platform team is also exploring how AI agents can be applied to AI governance review to speed it up and reduce manual work.
"We will be the CMDB for AI capabilities at the company," she said.
EY draws on AIOps for Now Assist rollout
A similar process unfolded with the business management consulting firm EY, which centralized most of its departments on a ServiceNow platform over the last 12 years, beginning with ITSM, IT operations management and HR service delivery for its 400,000 employees. EY, which is also a ServiceNow partner, began testing ServiceNow's Now Assist AI copilot in an ITSM environment eight months ago and rolled it out in production about four months ago. During that test process, the service desk team cleaned up its archive of knowledge articles by deleting old and irrelevant versions, and updated metadata categories and labels to ensure the AI copilot delivered accurate results.
Brian Eble, principal of the EY Technology team, said the service desk team also used previous versions of ServiceNow AIOps tools based on machine learning, which informed the AI copilot training process.
Everyone focuses on teaching AI things, but you also have to un-teach it things when new versions come out.
Brian EblePrincipal, EY Technology team
"That did help us with our change management and our transition to Now Assist, and allowed us to learn how the system learns, and how we can best teach it," Eble said. "One of the biggest things we learned as we went through that is that everyone focuses on teaching AI things, but you also have to un-teach it things when new versions come out. We've brought that forward as we've started to move to GenAI, which is much more easily adaptable."
The Now Assist copilot has begun to generate helpful results for EY's service desk, according to information Eble shared in a Knowledge 2025 presentation this week, primarily in resolution notes generation, incident and chat summarization, where it has been useful 70% to 80% of the time, as well as drafting 53 new knowledge base articles, each of which saved an estimated 30 minutes of time for human service desk agents.
Another ServiceNow customer, Bell Canada, has also seen early results from Now Assist in a call center environment, according to another Knowledge 2025 presentation.
"When our customer calls in, that AI capability is pulling in telemetry from our network … without requiring the [human call center] agent or the customer themselves to search for that," said Bruce Dean, senior vice president of field services for the telecom, during a Q&A at the end of that session. "We've saved about 500,000 calls so far."
Blue chips share plans for AI agents
EY has begun to test AI agents, once again starting in its ITSM environment.
"We have some targeted areas we want to look at in both HR and IT, but we'll start with the IT ones, just because that environment is further along right now," Eble said. "We know where there are opportunities for efficiency, and to create an end-to-end process. We've also begun to move RPA [robotic process automation] into ServiceNow, which is a big foundation for agents. That's usually the arms and legs, where the AI is the brain."
Aptiv PLC, a global automotive components supplier headquartered in Switzerland, plans to use AI agents with the real-time operating systems it acquired with Wind River Systems three years ago, according to a keynote presentation by Aptiv CEO Kevin Clark.
"ServiceNow helped [Wind River] operators detect network issues in real time and resolve them automatically before service is disrupted," Clark said. "Soon we're going to see the same end-to-end automation in vehicles, where ServiceNow AI agents are going to monitor systems running on Wind River and they're going to detect anomalies, like sensor degradation, for example, before they become problems."
ServiceNow officials presented demos of AI agent systems that are planned for use at multinational automaker Stellantis and biopharma company AstraZeneca this week. Those systems are based on adding AI agents to ServiceNow tools each company already has in production, but the agents aren't yet in production, according to a ServiceNow spokesperson.
In Europe, 85% of Stellantis' cars are scheduled and loaded for transport using AI, said Chris Taylor, group chief digital and information officer at Stellantis, during a keynote presentation this week.
In a separate keynote presentation focused on ServiceNow services management tools, Taylor said the company plans to use AI agents to link multiple dimensions of its manufacturing and business processes.
"We have AI agents in manufacturing right now that can respond to proactive information about a potential IT or OT [operational technology] fault and suggest remedial actions," he said.
Chris Taylor, left, of Stellantis joins Pablo Stern of ServiceNow on a keynote stage at ServiceNow Knowledge 2025.
ServiceNow AI roadmap refines measurements
Enterprises tend to measure the effectiveness and ROI of AI tools based on statistics such as calls deflected to a self-service portal. But one analyst said the industry must come up with better ways to measure AI as it matures.
"You deflected those calls, but how exactly did that produce value?" said Julie Mohr, an analyst at Forrester Research, in an interview this week. "Just deflecting calls doesn't always leave a satisfied customer. There needs to be metrics that help better understand the value of AI, rather than relying on older metrics that are from an age where we didn't have this technology."
Pablo Stern, executive vice president and general manager of technology workflow products at ServiceNow, argued in a separate interview that if customers remain unsatisfied, that shows up in further calls to human agents after the fact. This is already reflected in statistics about call volume measured in AI governance tools such as ServiceNow's AI Control Tower. But he conceded that AI metrics will likely evolve further.
"You'll be able to tie some of that back to hallucination and response rates, how much is escalating from an AI agent to a human, with a feedback loop that says, 'From these cases, the playbook that was created was not actually helpful,'" Stern said.
AI Control Tower metrics updates that reflect AI ROI in financial terms are already on the near-term roadmap, according to Nirankush Panchbhai, senior vice president of product management for the Now platform at ServiceNow.
"A CIO or CFO wants to see the business impact, in dollar amounts, that AI is creating," Panchbhai said in another interview. "The next step, which will be critical, is tying operational metrics to business value so they can start understanding how much was saved from the bottom line and how much in top-line opportunities was unlocked."
Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.
Dig Deeper on Systems automation and orchestration