NVIDIA and ServiceNow Launch Project Arc Enterprise AI Agent

At the ServiceNow Knowledge 2026 conference on May 5, NVIDIA and ServiceNow announced an expanded partnership to deliver governed autonomous AI agents to enterprises at two levels simultaneously: the employee desktop and the AI factory data center. The centerpiece of the announcement was Project Arc — an enterprise autonomous desktop agent capable of planning, writing code, executing commands, and adapting to unexpected outcomes across complex multi-step workflows, without requiring pre-built workflow templates for each new task.

Project Arc operates inside NVIDIA OpenShell, a sandboxed runtime environment that provides policy-based containment, auditability, and enterprise security for every action the agent takes. Every file read, command executed, and API called is logged and governed by ServiceNow AI Control Tower, which sets behavioral policies, monitors agent activity, and produces the audit trail that regulated industries require. Separately, ServiceNow completed its AI Control Tower integration with the NVIDIA Enterprise AI Factory validated design, extending the same governance layer to the infrastructure where large-scale model workloads run — connecting desktop-level agent actions to data center-level AI operations in a single managed environment.

The announcement reflects a critical inflection in enterprise AI: the shift from AI assistants that respond to prompts to AI agents that initiate, plan, and execute autonomously. While the previous generation of AI copilots helped humans do existing tasks faster, Project Arc is designed to take over entire task sequences — identifying the goal, breaking it into sub-steps, writing and executing the code to accomplish each step, and recovering from failures without human intervention. At ServiceNow Knowledge 2026, additional announcements reinforced this direction, including a new Accenture forward-deployed engineering program aimed at taking agentic AI from enterprise pilot to production at scale.

Gulf enterprises in sectors subject to formal regulatory oversight — banking, insurance, utilities, and government — have until now faced a dilemma: AI agents offer transformative automation potential, but operating them in regulated environments without auditable controls exposes organizations to compliance risk. The NVIDIA OpenShell and ServiceNow AI Control Tower architecture directly addresses this, providing the hardware-level sandboxing and workflow governance layer that transforms AI agents from research projects into production-grade enterprise infrastructure. UAE and GCC organizations investing in AI governance frameworks now will find this architecture provides a ready implementation path.

MawjazAI, Diverge's agentic AI platform, is built for precisely this operating environment — deploying AI agents that function autonomously within enterprise workflows while remaining fully auditable, policy-constrained, and aligned with organizational governance requirements. The Project Arc announcement validates the core design principle that Diverge has applied to the GCC market: autonomous AI capability and enterprise-grade governance are not trade-offs but requirements that must be achieved simultaneously for agentic AI to earn production trust.

The race to govern autonomous AI agents is now as strategically important as the race to build them. As AI agents gain the ability to write and execute code, interact with live APIs, and modify enterprise workflows in real time, the governance infrastructure organizations build today will determine which AI deployments scale safely — and which generate liability. The NVIDIA-ServiceNow partnership positions both companies as the preferred enterprise layer for the agentic AI era, and the organizations that adopt governed agent architectures early will establish the operational precedents and policy frameworks that define responsible enterprise AI in practice.