How ServiceNow uses internal AI pilots to launch successful customer tools

When Chris Bedi took the helm as ServiceNow's chief digital information officer in September 2015, he led a modest team dedicated to advancing artificial intelligence and machine learning. Fast forward to 2023, and ServiceNow is now actively piloting generative AI applications internally, exploring fifteen innovative ways to automate repetitive tasks for its employees. These internal projects are not merely experimental; they are strategically designed to set the stage for the company's AI offerings aimed at its client base. In May 2024, Bedi transitioned to the role of chief customer officer, with Kellie Romack, who joined in 2022 as senior vice president of digital technology experiences, stepping into the CDIO position. This shift highlights ServiceNow’s commitment to developing AI tools internally before introducing them to customers. Romack emphasized the importance of creating a cohesive AI ecosystem within the organization. "My team's mission is to foster unity and build this internal AI framework," she explained. Their efforts focus on automating IT help desk requests and generating code for software developers, demonstrating the company’s commitment to improving its own operations before extending these innovations externally. According to Kate Smaje, a senior partner at McKinsey & Company, adopting an internal-first strategy for AI development can significantly enhance a company’s confidence and learning, especially when incorporating employee feedback. The generative AI tools tested internally by ServiceNow have directly influenced the external products rolled out throughout 2023 and beyond. Romack’s team of digital technology engineers collaborates closely with product and platform engineers to create customer-focused tools. This collaborative pipeline was crucial in developing Workflow Data Fabric, a solution designed to connect clients’ various systems and data using machine learning. However, during its development, engineers identified delays in the system's data processing, which were promptly addressed before the tool's launch to customers in October 2024. As part of the AI Chain Reaction series, which explores the implications of AI adoption, Romack also led the creation of a governance tool named AI Control Tower in early 2024. This tool was designed to monitor internal AI use cases and the adoption of large language models, with a focus on AI governance, tracking efficiency, and employee engagement. By December 2025, ServiceNow had amassed over 240 internal and external AI use cases, with nearly 3,000 customers utilizing its AI tools. One of the standout successes of ServiceNow's internal AI applications has been within its IT service desk. The addition of agentic AI capabilities in August 2025 led to the introduction of Autonomous Workforce in February 2026, which allows customers to resolve common IT issues—such as password resets—without human intervention. Romack acknowledged that AI deployments can often be fraught with challenges. For instance, when the company first implemented generative AI for customer support case summarization, early attempts were not always accurate. The feedback loop from employees helped refine the tool before its rollout to customers. Romack now employs this rapid feedback strategy for all new feature launches, stating, "I don't wait two weeks; we evaluate within 24 to 48 hours. With AI, we have access to real-time data." However, Smaje cautioned that a successful internal AI initiative does not always guarantee success in external markets due to differences in security protocols and training requirements. "Experimentation is just the beginning," she noted. "The real challenge lies in transforming those insights into reliable systems that customers can trust and effectively utilize at scale."