Skip to main content

ServiceNow and NVIDIA launch Arc enterprise AI agent

21 OCTOBER 2026·3 MIN READ·4 SOURCES

ServiceNow says Arc, a personal AI agent for everyday work, is now generally available following an expanded partnership with NVIDIA. The announcement is dated October 21, 2026.

ServiceNow and NVIDIA launch Arc enterprise AI agent

Key takeaways · 5

  • 01

    Arc is generally available, but ZDNET reports that it currently supports Mac while Windows support remains in development.

  • 02

    ServiceNow says Arc works across workflows, email, Teams and other productivity tools, and can identify tasks and follow-ups proactively.

  • 03

    The control model combines ServiceNow’s Super Intelligence Control Tower with NVIDIA OpenShell; AI Passport can convey business rules for approving or blocking actions.

  • 04

    Review data handling before deployment: ServiceNow says Arc data may move to centralized environments, and customers can opt out of future collection of inputs, outputs and edits.

  • 05

    ServiceNow advises customers to test Arc, use human oversight and avoid relying solely on AI outputs for decisions.

What ServiceNow announced

ServiceNow says Arc is a personal, long-running agent intended to handle everyday work.[1] It can work across ServiceNow workflows, email, Teams and other common productivity tools, and can plan tasks across enterprise systems, execute them, check its work and adjust.[1] ServiceNow says the agent can also identify assigned tasks, pending approvals and follow-ups in employees’ backlogs, inboxes, meetings and conversations.[1] If an integration is missing, Arc can build one, according to the company.[1] The launch follows ServiceNow’s May 5, 2026 announcement of Project Arc, described as an autonomous desktop agent for complex work across enterprise tools and systems.[3]

How the agent is governed

ServiceNow says Arc is governed and secured by its Super Intelligence Control Tower, with enforcement through NVIDIA OpenShell.[1] Its AI Passport carries business rules—including risk tolerance and permitted external services—so NVIDIA enforcement layers can approve or block agent actions.[1] ServiceNow says Arc can work independently while the enterprise retains control, including granular control over which systems agents access, which APIs they call and which data they read.[1] The company’s stated aim for the expanded collaboration is to provide the trust enterprises need to scale autonomous AI.[1] For buyers, the practical question is how these controls map to their own policies and workflows.

Deployment, models and data

Arc can run on its own computer inside a secure NVIDIA OpenShell runtime or on an employee’s desktop, according to ServiceNow.[1] The company says the agent builds memory and develops skills as it works with an employee.[1] ServiceNow says Arc is designed to use local models for everyday work and frontier models for more demanding tasks, a design it says can lower inference costs and keep more data on the local device.[1] However, ServiceNow also says Arc data may be transferred from customer instances to centralized ServiceNow environments, potentially in another region or through a third-party cloud provider.[4] It collects Arc inputs, outputs and edits to improve its technologies, and customers can opt out of future data collection.[4]

Availability and adoption caveats

ServiceNow says Arc is generally available, while ZDNET reports that the application is available on Mac and that Windows support is still in development.[1][2] ServiceNow notes that some AI products and features are unavailable in specified restricted data centers, self-hosted environments or other restricted environments.[4] ZDNET reported that security controls are a major obstacle to wider business adoption of autonomous AI systems, following a summer marked by agent-led breaches.[2] ServiceNow advises customers to test and evaluate Arc, maintain human oversight and avoid relying solely on AI outputs for decisions.[4] Teams should verify platform access and data-handling requirements before planning a rollout.

Arc brings autonomous task execution into enterprise workflows, making its access rules, data flows and platform support practical deployment questions. Professionals evaluating it should validate those conditions and design human review into the work rather than treating agent output as a final decision.

Why it matters
Story quiz

Bring this to your team — one session a day, built for your industry.

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

For teams

How this developed

  1. 21 October 2026

    ServiceNow and NVIDIA launch Arc enterprise AI agent

Sources

AI fluency, one session a day, built for your work.