ServiceNow Says Agentic AI Lifted HR Capacity and Automated Support Work

ServiceNow executives described agentic AI changing work through measurable operational shifts: HR business partners moving from serving about 400 people to about 1,000, commission-status answers falling from four days to eight seconds, 90% of IT service-desk work moving from first touch to resolution autonomously, 85% of support-desk staff redeployed into other roles, and 91% of customer support handled through self-service and AI. Those figures came from Jacqui Canney and Kellie Romack, who described ServiceNow’s own internal deployments in a conversation sponsored by ServiceNow.

Alex Kantrowitz opened the HR example with a specific claim: ServiceNow had used AI to help HR business partners move from serving about 400 people to serving about 1,000. Canney answered by describing the broader people-operations work behind that shift. ServiceNow, she said, asked employees for ideas about where AI could improve work and received roughly 1,000 use case suggestions “in short order.” Those ideas then went through prioritization and governance before agents and AI were deployed into the people-operations space.

The result, Canney said, was that the function could serve a much larger employee base “without having to add headcount.” ServiceNow had grown from about 14,000 employees to almost 30,000, while her team served double the number of people. The claim was not that AI eliminated whole HR jobs. Canney drew the distinction carefully: AI took over parts of jobs and reallocated capacity.

That distinction mattered because Kantrowitz pressed on the familiar line that AI automates tasks rather than jobs. Canney’s answer was that the company did not tell people, in effect, “I don’t have this job anymore, but I have this job now.” Instead, it tracked the capacity created by automation and moved people into adjacent work: building more use cases, training AI, and handling higher-order cases that still needed more human involvement.

Canney also emphasized the operating environment around the technology. Her team needed to feel secure enough to try the tools, fail in limited ways, and then pursue more return once a use case showed traction. She described the “safety of trying” as part of why the work succeeded.

Romack’s most concrete early example was commissions. ServiceNow’s finance organization had been “drowning in ticket requests” from sales employees who needed to understand where they stood on commissions. The old process took four days. After ServiceNow reimagined the work and applied AI with security controls, Romack said the answer could be delivered in eight seconds.

Romack framed the commission use case as a win across groups: finance employees could move to higher-value work, sellers could spend more time with customers, and employees got faster answers about compensation. Kantrowitz noted that this was a more ordinary but consequential version of AI automation than the public imagination often reaches for. Instead of replacing salespeople with AI salespeople, automation removed an internal process that pulled humans away from planning, customer work, and compensation design.

Canney used the same example to make a broader organizational point. AI, in her view, “disintegrates the org chart” because the problem is not contained inside HR, finance, or operations. A commission question may touch all of them. The implementation lens shifts from automating a department’s existing task to solving an employee-facing workflow across functions.

Kellie Romack rejected the idea that an AI program should begin by layering AI on top of an old process. Her formulation was direct: “Don’t automate the old, reinvent to the new.” In the ServiceNow examples, that meant stepping back from the existing ticket flow and asking what the business outcome should be, rather than simply making the old queue move faster.

That approach also changes the skills conversation. Jacqui Canney said the company starts by redesigning the work with AI embedded, then asks what skills are needed to perform the redesigned work. Upskilling is not treated as a parallel HR initiative but as part of the operating model that makes the automation useful.

ServiceNow University was one mechanism for that shift. Romack said she and Canney had sketched the idea years earlier as a place where people could learn and grow, and that partners, customers, employees, and even people she knows outside the company use it. She described the training as covering basic AI skills, prompting, efficiency, reimagining work, design thinking, and other material. Canney added leadership to the list and said the program had 1.8 million learners, with a goal of reaching three million.

Alex Kantrowitz challenged the common weakness in corporate training programs: employees may learn a tool, return to their desk, become more productive, and then collide with managers who do not know how to handle the new capacity. The bottleneck, in his view, is not only education but permission.

Romack said ServiceNow tries to create that permission through practices such as “AI Required Workdays,” where employees are expected to invest time and energy in using the tools. Her role, she said, is to infuse AI into the work rather than forcing employees into “swivel chairs” moving between disconnected technologies.

Canney put the emphasis on managers. She called managing teams one of the hardest jobs in the current environment because managers must deliver their own work, help teams build capacity, and create safety for experimentation. ServiceNow invests in managerial skill-building and forums where managers can exchange practices, whether they are new managers, new to ServiceNow, or managers of managers. She argued that this is not only a ServiceNow issue; managerial capability is a major unlock for AI adoption in any company.

Romack tied the adoption claim to pilots. ServiceNow runs pilots, learns from them, expands the groups, and then drives broader adoption. She said this approach has produced 95% AI adoption across the company.

Alex Kantrowitz raised the more radical version of the AI-management argument: some in tech suggest companies may not need middle managers at all. In that scenario, individual contributors would report what they are doing, the CEO’s agent would ingest the data, and leadership could synthesize decisions without a conventional management layer.

Kellie Romack did not dismiss the possibility that structures will change, but she called the issue “more complicated than that.” ServiceNow’s own IT service desk became her example. The desk supports about 30,000 people in the organization. Romack said the company applied an autonomous workforce model to that function and moved 90% of everything coming through the IT service desk from first touch to resolution autonomously with AI.

That did not leave the same service desk staffed in the same way. Romack said ServiceNow moved 85% of the people who had been doing that support work into higher-value jobs, including security operations, AI operations, and executive briefing centers. She described 15% as what remained on the IT service desk, with the work itself transformed: those employees help with complex cases, monitor what agents are doing, improve the organization’s systems, and manage the agentic workforce.

Kantrowitz summarized the shift as “managing the agents.” Romack agreed, but added “and people.” The remaining human layer still escalates complex work, collaborates across the organization, builds relationships, and supplies context agents do not have.

Jacqui Canney focused on judgment. Agents may get smarter and work more closely with humans, she said, but leaders still decide what the organization values, which use cases are worth doing, and what should be done rather than merely what can be done. That choice requires pattern recognition, experience, and human judgment.

She also described workforce design as an open decision rather than a predetermined outcome. Companies may become more like an hourglass, a pyramid, or some other shape depending on where AI is deployed and what governance is imposed. Kantrowitz explained the hourglass as a model with leadership at the top, slimmer middle management, and many people underneath; he also mentioned a “diamond” model, with fewer people at the top, many decision-makers in the middle, and fewer entry-level roles because AI absorbs some of that work. Canney’s point was that these shapes are now choices management and CEOs must make.

Alex Kantrowitz put the harder labor question plainly: many people hear “redeploy today” and assume it means “layoff tomorrow.” Kellie Romack answered that ServiceNow handled the support-desk shift through conversations with the people affected, looking at their career aspirations, skill sets, flexibility, and adaptability before making choices about where they could move.

Some moved into AI operations. Some moved into security. Romack framed those moves as career expansion rather than simple reassignment. “Sometimes growing your career means getting wider,” she said — learning more, building experience in adjacent domains, and moving into work with more long-term value.

She also noted the practical demand side: as a CIO, she said, there is no shortage of backlog. The opportunity, in her description, is to use AI to clear lower-value support work so technology teams can serve more people and more products across the company.

Jacqui Canney added that ServiceNow assessed the whole company on AI capability, but not as a threat. The assessments were role-specific rather than uniform across all employees. A third party helped determine where people sat on relevant skills, then ServiceNow pushed personalized training to help them improve. That gave leaders, including Romack, an “x-ray” of AI capability across teams: where strengths already existed, where more training could create growth, and where conversations were needed about whether an employee was willing to try a new path.

This is also where Canney’s view of career progression departed from the old level-by-level corporate ladder. She said it would be too assumptive to say everyone wants a linear career path, or that everyone wants an adaptive one. Employers need to be transparent about what the company is building and what it needs. The workforce that wins, in her formulation, is adaptable and agile.

Canney said that when ServiceNow hires people who are “AI native,” the company can teach them the enterprise. Her broader point was that some enterprise knowledge can be taught more easily than adaptability, acuity, and the human traits employers increasingly need. Technology speed alone is not enough; management has to create the conditions for people to move with the work.

Kellie Romack said ServiceNow measures AI work through value from the beginning. Each internal AI use case starts with a value case, is built against that value, and is evaluated on the value it creates. From the technology perspective, her job is to create value, streamline work, reduce friction, and protect employees and customers.

The core mechanism she described is an AI control tower. Romack said ServiceNow can see how many AI agents are running at any point in the day, including both ServiceNow and third-party agents; how people are using them; adoption; relevant metrics; value created; and longevity. The same system provides governance and security visibility.

ServiceNow built the control tower out of necessity. The company was building agents and deploying AI internally, and Romack concluded it needed a way to see and manage the environment. After using it internally, ServiceNow productized it for customers. Romack said the company had been using it for about a year.

Jacqui Canney added that this kind of visibility prevents duplication. Without it, multiple teams might build versions of the same agent without knowing it, creating unnecessary cost and fragmentation. With visibility, teams can see that a capability already exists elsewhere or combine efforts into a stronger use case.

Romack’s phrase for the cost issue was concise: “tokens are a new currency.”

The control problem becomes sharper because AI development is not limited to central IT. Romack distinguished between “official IT,” which she leads, and citizen development, which ServiceNow has long had. She said the “gray zone” between citizen development and core IT is growing. Her team is trying to preserve flexibility for business teams, including Canney’s HR organization, while still maintaining safety, security, and governance.

Canney gave an example from HR. An HR business partner was often asked for benchmarking data, such as whether the size of an organization was appropriate relative to industry benchmarks. Previously, that might require calling a consultant or doing substantial research. Because Romack’s team had agents operating in a governed environment with access to approved external data, the HR partner could build an agent to answer the question. Canney said that use case became valuable enough that it was being made part of ServiceNow’s product.

The example illustrated the tension both leaders kept returning to: business users should be able to innovate close to the work, but not by running unmanaged agents across sensitive company data.

Alex Kantrowitz asked how a company goes from thinking an agent might help with IT tasks to trusting it to resolve 80% or 90% of queries. Kellie Romack answered through customer support, which she described as ServiceNow’s most important information because it touches customers directly. She said 91% of ServiceNow customer support is handled through self-service and AI.

That level of automation was not a switch-flip. Romack described a stair-step process over roughly 18 months: monitoring, keeping humans in the loop, then moving humans “on the loop,” where they monitor and manage agents rather than handle every case directly.

Jacqui Canney said content quality is central. Her advice for companies getting started was to spend time on content, then begin in areas where imperfect answers are tolerable while the content and system improve. The safe starting point matters: a company needs an audience and workflow where learning can happen without excessive risk.

Romack said this is why ServiceNow experiments internally and runs pilots frequently, especially between technology and HR. Canney characterized the relationship as brave but not perfect: the teams give each other grace, learn together, and iterate.

Romack’s final leadership advice was to move from “a black box to a glass box.” Leaders need to understand what is being automated in depth so they can know whether it is being automated correctly. Canney’s advice was cultural: leaders should model curiosity, ask good questions, and keep the organization safe for experimentation rather than governed by fear of breaking something.