WHITE PAPER AI value framework How ServiceNow measures the value of AI WHITE PAPER Table of Contents AI agents are supporting 5 personas .........................................................4 Thinking about measuring value differently in the world of AI ......... 5 How we measure the value of AI for: ........................................................... 5 End users ...................................................................................................................5 Human agents .......................................................................................................9 Process owners ..................................................................................................... 15 Developers ............................................................................................................. 19 Leaders ................................................................................................................. 20 Self-service efficiency score ..........................................................................21 Key takeaways ...................................................................................................23 ServiceNow works on ServiceNow. ServiceNow works on ServiceNow. We use our platform every day to drive speed and scale across the enterprise. And, as customer zero, our Now on Now practitioners are the first to adopt our products. This allows us to support our customers by sharing our learnings. We’re pleased to share with you our approach to measuring the value of AI. 3 WHITE PAPER Looking at time spent—how much and by whom—is a more useful productivity metric to trace the impact of AI. Although we tend to think about technology in terms of new gadgets or inventions, most of the productivity improvements in history came down to workers gaining the ability to do some things faster, freeing up more time for other activities. In short, process innovations are more important than product innovations. “Why It’s So Hard to Measure AI’s Effects on Productivity” Diane Coyle, Bloomberg.com March 20, 2025 4 WHITE PAPER Figure 1: ServiceNow AI capabilities support 5 personas AI agents are supporting 5 personas AI starts with data and ends with experience. AI will transform how organizations unlock efficiency and increase productivity. At ServiceNow, we have implemented more than 500+ AI use cases across five personas, accelerating self-service and minimizing human agent intervention on low-complexity requests. Infusing AI into search, conversations, and workflows is enabling requestors to quickly find solutions, leading to higher self-service rates. ServiceNow AI agents can solve issues faster and be more productive because AI is solving for more low-complexity issues and automating tasks. Our developers can code faster on the ServiceNow Platform. And our leaders have comprehensive, deeper, and faster insights to inform decision-making. In short, AI is changing the way we work by improving the speed at which we work, the experience we have while working, and the decisions we make to protect and grow our business. With purposeful automation across ServiceNow, our employees can be more productive and focus on more strategic and creative work. 5 WHITE PAPER Thinking about measuring value differently in the world of AI Using traditional metrics to measure AI’s impact will not provide an understanding of the transformative impact AI can have on departments, operations, and enterprises. Simply put, we need to think about measuring value differently in the world of AI. At ServiceNow, we measure AI’s impact by calculating the value of its overall boost to productivity. We call this our AI value framework. The AI value framework measures value through usage, user acceptance, and productivity time value expressed as total hours. Hours can be interpreted as productivity gained or cost takeout, depending on the customer and persona where value is associated. We calculate the financial impact of AI by measuring productivity gain in time saved, requests avoided or improved growth and, where needed, use reasonable and defensible assumptions. AI value for end users We measure the productivity value on end users’ (or requestors’) based on: 1. AI Search productivity gain 2. Conversational AI driving productivity gain 3. Requestor time saved with successful automated requests 4. Human agent time saved with successful agentic workflows 1. AI Search productivity gain Simplified search is a powerful tool. At ServiceNow, AI Search and Now Assist Search are driving productivity gains by enabling requestors to find answers quickly and by reducing live help costs on low-complexity issues (See Figure 2). HOW WE ME ASURE Figure 2: Example of AI-generated search results Value calculation for AI should be tied with actions to transform the operating model of the business - skills, hiring, technology and processes. Vijay Kotu Chief Analytics Office, ServiceNow 6 WHITE PAPER Figure 3: Measuring value of time saved due to successful AI Search EXAMPLE We define a successful search as when an employee resolves their issue or finds an answer without the need for human assistance by clicking on a Search Result or viewing Now Assist Search summary result. We measure the value of AI Search and Now Assist in AI Search both for the requestor who has resolved their issue, and the human agent who has saved time by not needing to intervene. AI Search and Now Assist in AI Search are available in our employee portal and through conversations with our Virtual Agent. Both offer value to requestors by helping them find information and resolve issues. We calculate the value by multiplying the number of successful searches by the time saved per interaction compared to traditional click-through searches (see Figure 3). Here is our approach: • AI Search in our employee portal and facilitated by Virtual Agent: Our analysis shows that the time saved per AI Search in our portal or with Virtual Agent is 2.5 minutes. We calculate this based on the assumption that the maximum time spent in search is no more than 5 minutes. We measured that a requestor spends approximately 2.5 minutes clicking through results on AI Search, so we use that as a constant figure of time saved per AI Search. We multiply that by the number of successful searches as well as an estimate of hourly savings. • Now Assist in AI Search: We measure the impact of GenAI on our search success. Again, we assume the maximum time spent on any search is 5 minutes. The average amount of time spent engaging with Now Assist Search is 30 seconds. This means the time saved per search with Now Assist is 4.5 minutes—2 minutes faster than AI Search (without GenAI). We then multiply this savings of 4.5 minutes by the number of successful searches with a Now Assist result as well as an estimate of hourly savings. *A search is considered successful when an employee resolved their issues or finds an answer without the need for human assistance 7 WHITE PAPER 2. Conversational AI productivity gain Agentic AI can automatically execute intelligent workflows. At ServiceNow, we use agentic AI within an AI-powered workflow with a virtual agent interface where a requestor engages to resolve their need. Agentic AI executes pre- configured topics and conversational catalog requests to enable successful self-service and avoid the need for live support. Agentic AI will further enable self-service efficiency and reduce the amount of live support help needed. Agentic AI within a conversational virtual agent session is considered successful when a requestor resolves their issue or finds an answer without the need for live support. A successful session – via Agentic AI, topic conversation or conversational catalog – can save requestors time by helping them execute an automated workflow. Figure 4 outlines how we approach calculating the value of requestor time saved through successful conversational AI: *A virtual agent sessions is considered successful when an employee resolved their issues or finds an answer without the need for human assistance EXAMPLE Figure 4: Measuring value of requestor time saved due to successful virtual agent conversation • We start by calculating the total successful virtual agent sessions: We define a successful virtual agent session as the number that did not require human agent support. • Total hours saved in self-service using topics and conversation catalog: Our data indicates that employees spend an average of 3.68 minutes on a successful virtual agent topic or conversational catalog conversation. This would otherwise have become a low-complexity case, requiring an employee to spend around 15 minutes per issue. Accordingly, each successful virtual agent conversation saves the requestor an average of 11.32 minutes. • The value of time saved: We then translate the value of this requestor time saved by multiplying it by employee costs per hour. This gives us the value of requestor time saved because of conversational AI experiences. 8 WHITE PAPER 3. Requestor time saved with successful automated requests Agentic Workflow automation refers to specific tasks previously performed entirely by humans that are now getting automated and augmented. AI is augmenting customer service human agent tasks by automating case workflows, augmenting case resolutions, providing case context using summarization, generating resolution notes, and automating knowledge base articles. A successful automated request can save time for the requestor who does not have to open a case to get live help. We look at the number of agentic AI request items in the service catalog which were successfully completed (See Figure 5). We know these request items would have required live help because they were in the service catalog. This would have required an employee to spend around 15 minutes per issue. Instead, agentic AI request items save the employee 15 minutes. We then multiply this by the employee cost per hour to give us the value of requestor time saved.
