This article outlines key developments and expectations in the rapidly evolving world of AI—particularly as it is being embedded and operationalized within the ServiceNow platform. By now, most professionals have interacted with some form of Generative AI—be it Co-Pilot, ChatGPT, Gemini, LLaMA, DeepSeek, or others. Many have also activated NOW Assist within ServiceNow and are exploring its Generative AI features, accessible through Virtual Agent, the Employee Self-Service portal and Fulfiller Workspaces.

AI at Work: How ServiceNow is Transforming Enterprise Automation with AI Agents

The newest innovation entering the scene is a suite of “AI Agents”—autonomous entities that can reason, decide, and act, depending on their configuration. ServiceNow is introducing a structured AI Landscape comprising Triggers, Communication Agents, Orchestrators, Work Agents, and Tools. The orchestrators and agents are configured using AI prompts, meaning the quality of their performance is directly tied to the quality of those prompts.

These AI Agents bridge the gap between large language models (LLMs) and traditional scripts, workflows, and business logic. However, it’s important to note: existing tools don’t magically become more capable simply because AI is now involved.

What’s Coming in 2025

In May 2025, ServiceNow will launch an upgraded, AI-powered version of its chatbot—branded “NOW Assist Windows Experience”—available on instances running Xanadu Patch 7 or later. The interface will closely resemble ChatGPT’s familiar, intuitive web experience, moving away from the classic pop-up design. Users can also expect the bot to retain and learn from previous interactions, offering a more personalized and seamless experience.

A significant addition will be Retrieval-Augmented Generation (RAG), which enables the platform to enhance responses by combining LLM outputs with enterprise-specific content like knowledge articles. Conversations—once anonymized and summarized—can be securely routed to external models, enriched with relevant content, and returned with contextually aware answers.

The Bigger Picture

Tech giants such as Dell, Apple, and Indian IT leaders like TCS, Infosys, Capgemini, Wipro, HCL, Tech Mahindra, and Cognizantare working on proprietary LLMs containing enterprise-relevant data. In line with this, ServiceNow plans to roll out dozens of connectors throughout 2025—enabling integration with platforms like SharePoint, Confluence, Google Drive, SAP Document Management, and ServiceNow Docs. This means enterprise knowledge doesn’t have to live solely inside ServiceNow to be usable.

To address data privacy, personally identifiable information (PII) will be carefully filtered before any external AI processing. Looking ahead, ServiceNow aims to retain conversation context longer and even reintegrate filtered data into authorized responses where appropriate—offering both security and continuity.

Evolving Ecosystem

Major players including SAP, Google, Salesforce, Microsoft, and ServiceNow are all crafting unique interfaces and AI infrastructure for enterprise use. Each will have their own orchestrators and AI agents—some connected to proprietary models, others to public ones. The trend points to growing interoperability, potentially reducing the current fragmentation across user interfaces. Still, it’s unlikely that enterprises will standardize on a single AI platform.

A notable effort in this direction is an Agent-to-Agent (A2A) protocol, under development by a consortium that includes Google and ServiceNow, aimed at enabling discovery and collaboration among thousands of AI agents.

MS Teams and Beyond

Another leap forward: ServiceNow is integrating the NOW Assist CoPilot Engine Agent with Microsoft Teams. This offers a robust alternative to traditional virtual agents, leveraging real-time AI conversations that access the ServiceNow Catalog and Knowledge Base. If needed, complex interactions can be escalated to human agents in Service Operations Workspace, where issues can be resolved with the help of AI-enhanced fulfillment tools.

Organizations can configure, test, and govern these agents directly on the platform. ServiceNow will deliver 15 out-of-the-box (OOTB) agents in Q1 2025, growing to 50 in Q2, and hundreds by year-end. Recognizing that not all customers immediately upgrade to the latest release, ServiceNow intends to backport some AI agents to earlier versions.

Key Takeaways

  1. Information Access is Changing
    In the past, enterprise models had to be explicitly trained. Increasingly, they retrieve relevant data from external sources in real time. RAG provides a hybrid model: combining external answers with internal insights—saving time and improving accuracy.
  2. No One-Platform Reality
    Enterprise AI will not consolidate into a single platform. Expect multiple AI-powered platforms, each with distinct roles and audiences, interacting as needed. Organizations must plan to invest in AI capabilities across several ecosystems.
  3. Strategic Decision-Making is Required
    Enterprises should assess which use cases offer the most value, what resources are needed, and how best to build an AI roadmap. A key decision: whether to build custom AI agents, clone ServiceNow’s OOTB agents, or source from partners and managed service providers.