ServiceNow: Internal AI, Key to Innovative Customer Solutions

An Internal-Focused AI Strategy for Tangible Customer Results

ServiceNow has adopted a unique approach to developing its artificial intelligence tools, prioritizing internal testing before offering them to clients. Since 2023, the company has implemented over 240 AI use cases, with a deployment goal set for December 2025. Kellie Romack, Chief Digital Information Officer at ServiceNow, detailed this ambitious timeline.

When Chris Bedi joined ServiceNow as Chief Digital Information Officer in September 2015, he led a small team dedicated to AI and machine learning. In 2023, the company began testing generative AI applications internally, exploring 15 ways to automate employees' repetitive tasks. These pilot projects, initially intended for internal use, were designed to pave the way for future AI offerings from ServiceNow for its clients.

In May 2024, Bedi transitioned to the role of Chief Customer Officer, while Romack, who was recruited in 2022 as Senior Vice President of Digital Technology Experiences, became CDIO. This transition underscores ServiceNow's strategy of developing its AI tools internally before deploying them externally. Romack explained to Business Insider that she leads efforts to create and implement AI tools capable of automating IT help desk requests and generating code for developers.

The Importance of Internal Feedback for Successful Development

Kate Smaje, a senior partner at McKinsey and Company, emphasized that the internal approach adopted by ServiceNow allows companies to build confidence and learn by soliciting feedback from employees. The generative AI tools tested and deployed internally have had a significant impact on the external products launched in 2023 and beyond.

ServiceNow's digital technology engineers, under Romack's leadership, work closely with product and platform engineers to develop tools for clients. This process has enabled ServiceNow to create Workflow Data Fabric, a tool that connects disparate client systems, data, and employees through machine learning.

During the development of this tool, some engineers noticed that the system took too long to send data. This issue was resolved internally, and Workflow Data Fabric was made available to clients in October 2024.

Governance Tools for Responsible AI

In the first quarter of 2024, Romack also led the development of a governance-focused tool called AI Control Tower. This tool allows for tracking internal AI use cases and the adoption of large language models. When this product was launched for clients in May 2025, ServiceNow focused on three main themes: AI governance, tracking efficiency gains, and employee adoption.

By December 2025, ServiceNow had over 240 internal and external AI use cases, with nearly 3,000 clients using its AI tools. Romack mentioned that one of the most successful internal applications of generative AI is found in its IT help desk, where agentic AI capabilities were added in August 2025. This project led to the launch in February 2026 of Autonomous Workforce, a tool that enables clients to resolve common IT issues without human intervention.

The Challenges of AI Implementation

Romack acknowledged that deploying AI is not always a linear process. For example, during the initial use of generative AI for customer support synthesis in 2023, the early results were not always accurate. ServiceNow had to refine its tools using employee feedback to identify necessary adjustments before offering them to clients. Today, Romack applies this strategy to every new feature.

She emphasized the importance of reacting quickly: "I don't wait two weeks," she stated. "We talk within 24 to 48 hours of reviewing it. With AI, we have real-time data."

However, as Smaje noted, an AI tool that enhances internal productivity does not always translate directly for external clients due to differences in security protocols and employee training. "Experimentation is just the first step," she asserted. "The real challenge, and where value is created, is transforming these learnings into robust systems that clients can trust and use at scale."