Artificial Intelligence at ServiceNow

ServiceNow is an American enterprise software company headquartered in Santa Clara, California. The company employs more than 29,000 people globally and reported fourth-quarter 2025 subscription revenue of $3.57 billion, with fiscal 2026 subscription revenue guidance of $15.53 billion to $15.57 billion.​

The company has also invested heavily in AI and automation to improve workflow efficiency and enterprise productivity. ServiceNow acquired Passage AI to strengthen conversational AI capabilities, expanded its partnership with NVIDIA to support autonomous AI agents, and committed $1 billion through its venture arm to back enterprise software and AI-related startups.

ServiceNow has also made a CA$110 million investment to support AI adoption in Canada’s public sector, including infrastructure and an AI Center of Excellence.​

The company is actively using its own AI platform (Now on Now) to achieve significant ROI, demonstrating how C-suites can move beyond AI experimentation into tangible, scalable automation.

Here are two key AI use cases, leveraging Now Assist (GenAI), that enterprise leaders can learn from:

Reducing agent documentation time with embedded generative AI: Leveraging generative AI inside existing ITSM (IT Service Management) and CSM (Customer Service Management) workflows to automate summaries, resolution notes, and knowledge article creation, helping agents save time and focus on higher-value support work.

Predicting customer escalations before they happen: Using machine learning and real-time event monitoring to identify at-risk accounts early, automate proactive outreach, and reduce costly customer escalations.

Reducing Agent Documentation Time With Embedded Generative AI

​A San Jose State University research paper, “Empowering customer service with generative AI”, documents that customer service agents spend 35-45% of their time on repetitive documentation, creating $2.6 billion in annual U.S. labor inefficiency across enterprises.​

A Harvard Business School study on specialization in repetitive work similarly finds that service agents lose significant time on documentation and summarization, diverting focus from high-value problem-solving.​

To solve this, ServiceNow launched Now Assist for ITSM and CSM, embedding generative AI directly into agent workspaces per the company’s official ITSM Now Assist documentation and CSM Now Assist guide. Instead of standalone chatbots, it automates summarization of incident histories, drafting resolution notes, and generating knowledge articles within existing workflows.

Using machine learning and generative AI, the company automated the mundane aspects of support cases, such as summarizing long incident histories, drafting resolution notes, and generating knowledge base articles.

ServiceNow reports: in Now Assist, LLMs ingest case context, generate editable summaries/notes in seconds,, and provide agent reviews in record time, so there is no need for context-switching to external tools.

​

​Screen of chat showing conversation between the IT agent and Now Assist (Source: ServiceNow)​

In the same whitepaper, the company shared that Now Assist generates notes within seconds, allowing agents to review and refine them versus creating them from scratch. This reduced the time needed for each resolution note by approximately 80%.

​ServiceNow also shared that, on average, ITSM agents save 4-6 minutes per use, and CSM agents save 12-16 minutes per use. This proves that enterprise AI value comes from embedding generative AI into existing workflows, not standalone demos.

Predicting Customer Escalations Before They Happen

Historical monitoring relies heavily on manual checks of tickets and events, making it hard to spot deteriorating experiences before customers escalate. Without a scalable way to predict which accounts are at risk, proactive outreach remains inconsistent and often too late.​

ServiceNow turned a classic reactive support model on its head by using machine learning to predict and prevent customer escalations. According to a case study published by ServiceNow, instead of waiting for customers to complain or threaten to escalate, the team now leans on ServiceNow’s own Predictive Intelligence and Event Management capabilities to proactively identify at‑risk accounts and reach out before issues snowball. ​

The initiative is built on ServiceNow’s Predictive Intelligence framework, which hosts the underlying machine‑learning model, and Event Management for real‑time ingestion of performance‑related events.

Within Predictive Intelligence, a supervised model trained on historical escalation patterns analyzes tickets, surveys, CSAT scores, and engagement signals. Event Management adds real‑time system alerts.

How the workflow operates:

Builds and trains the model: Historical escalations and correlated events are transformed into structured features, which are then used to train and validate an XGBoost classifier through a PoC/PoV phase.

Deploys real-time risk scoring: Once live, the model continuously scores customers as new tickets and events arrive, assigning escalation-risk labels such as low, medium, or high.

Automates proactive intervention: When a customer moves into the high-risk category, ServiceNow workflows automatically generate priority alerts, assign follow-up tasks to support or account teams, and surface recommended playbooks with next steps.

Screenshot showing the benefits of this solution (Source: ServiceNow)

​

Over time, the outcome of each engagement, whether the escalation was averted or not, feeds back into the model, continuously refining its predictions. ​

It also lays out the business results as follows:​

Timelier interventions have led to faster response and resolution times, higher customer satisfaction scores, and smoother renewals and upsells, as at‑risk accounts are stabilized before they become vocal.

Before the model, only about 11% of customer engagements were proactive; after implementation, roughly 68% of engagements became proactive, enabling earlier and more systematic outreach to at‑risk customers.

The system helped engage hundreds of customers per year, preventing a large share of escalations while keeping the false‑positive rate around 3%, so engineering resources are not wasted on unnecessary cases.