Enterprises are shifting focus from AI development to its operational deployment. The challenge now is not just building intelligent systems but ensuring they can execute tasks autonomously while maintaining governance, security, and efficiency.

NVIDIA and ServiceNow have expanded their partnership to deliver specialized autonomous AI agents tailored for real-world enterprise workflows. These agents leverage NVIDIA's accelerated computing infrastructure—including the Blackwell platform—and open models, paired with ServiceNow's enterprise-grade workflow tools and governance frameworks.

  • What we know so far:
  • Project Arc: A persistent autonomous desktop agent designed for knowledge workers, including developers and IT teams, capable of continuous self-improvement.
  • Secure execution: Integration with NVIDIA OpenShell ensures agents operate within defined boundaries, preventing unauthorized actions or data exposure.
  • Customization tools: The NVIDIA Agent Toolkit provides Nemotron open models and domain-specific skills for building tailored AI applications.
  • Performance gains: Blackwell offers 50x greater token output per watt compared to Hopper, cutting costs by nearly 35x per million tokens.
  • Governance integration: ServiceNow's AI Control Tower works with NVIDIA Enterprise AI Factory to manage large-scale agent deployments, ensuring auditability and compliance.

The partnership addresses critical enterprise needs—open, customizable models; domain-specific expertise; and robust security measures. Project Arc integrates natively with the ServiceNow AI Platform via Action Fabric, embedding governance into every workflow interaction.

Efficiency is a defining factor in scaling AI agents across vast enterprise environments. Blackwell's performance leap over Hopper—50x more tokens per watt and 35x lower costs—could accelerate adoption from pilots to production. However, questions persist about adaptability: Can these agents handle the complexity of diverse workflows without compromising security or performance?

The collaboration marks a turning point in enterprise AI, moving beyond theoretical potential to practical deployment. Success will hinge on whether the combined infrastructure can balance speed, governance, and real-world utility—positioning early adopters ahead in the race for AI-driven operational excellence.