Nvidia and ServiceNow debut advanced AI agent for enterprises

ServiceNow and Nvidia have partnered to develop a new AI model specifically designed to power enterprise-grade AI agents that can autonomously perform tasks. Unveiled at ServiceNow’s Knowledge 2025 conference, the Apriel Nemotron 15B aims to deliver the sophisticated reasoning capabilities needed for agentic AI while maintaining smaller model size, lower latency, and reduced costs. This development addresses a critical need as businesses increasingly seek to integrate AI agents into their operational workflows to automate complex processes.

The big picture: ServiceNow and Nvidia have launched Apriel Nemotron 15B, an open-source reasoning language model that delivers advanced AI capabilities in a more efficient package.

• The model was trained using Nvidia Nemo, the Nvidia Llama Nemotron Post-Training Dataset, and ServiceNow’s domain-specific data.
• According to benchmark testing, the model delivers promising results for its size category, positioning it as an effective foundation for agentic AI workflows.

Why this matters: Reasoning capabilities are crucial for AI agents that need to autonomously perform tasks across various enterprise settings without human direction.

• The model’s smaller size makes it less expensive and faster to deploy while still delivering the enterprise-grade intelligence necessary for complex business applications.
• The efficient design allows it to run on Nvidia GPU infrastructure as an Nvidia NIM microservice, reducing the computational resources required.

Key innovation: Beyond the model itself, the companies unveiled a joint data flywheel architecture that creates a continuous improvement loop for AI systems.

• The architecture integrates ServiceNow Workflow Data Fabric with select Nvidia NeMo microservices to capture and utilize enterprise workflow data.
• This system incorporates necessary guardrails to protect customer data, ensure secure processing, and maintain customer control while enabling the creation of highly personalized, context-aware AI agents.

Behind the technology: The data flywheel approach represents a strategic shift in how AI models can improve over time through real-world usage.

• As AI agents complete tasks within enterprise workflows, the data generated feeds back into the system to refine the models’ reasoning capabilities.
• This continuous learning process allows the AI to become increasingly adept at handling complex business scenarios within specific organizational contexts.