Introducing ServiceNow Arc: Your Personal Assistant, Governed for the Enterprise

[Paty Montesinos]

Introducing ServiceNow Arc: Your Personal Assistant, Governed for the Enterprise

Your most important work is scattered, and your AI can't keep up

Think about a typical morning. You check overnight alerts, pull up emails that need a response, review today's meetings, and copy details from one system to another. Most of that work happens between applications, and it's manual. Workflows are great at the steps we design for, but a lot of real work falls outside the ServiceNow platform, across the apps and enterprise systems your team uses every day: the tasks, the handoffs, and the "go find this and bring it back" moments that nobody has automated.

Knowledge workers spend much of their day on the work around the work. The insights that matter most are spread across systems that don't talk to each other. A high-priority incident or a stalled approval sits in your ServiceNow instance; the context is buried in an email thread or a Teams chat; and, for technical teams, the rest of the story lives in GitHub. Pulling it together means a dozen browser tabs, a lot of copy-and-paste, and the nagging sense that something slipped through.

Many people have turned to AI agents for help, but those tools bring their own frustrations. Most lock you into a single model family, even when another model would do the job better or cost less. And even when you can switch, it's up to you to remember. You might reach for a lighter model for a quick summary and a more powerful one for deep analysis, but most people don't, so they overpay for simple tasks or underpower complex ones.

Meanwhile, IT and security leaders are stuck. Employees are already running powerful agents against email, customer records, and sensitive data, and CIOs and CISOs often can't see it. AI spend keeps climbing, and without a governed way to adopt these tools, leaders can't say yes to the AI their people want and need.

People need a personal assistant that finishes their work across every system they use. Leaders need one they can see, control, and trust. Until now, you had to choose.

Meet Arc: an assistant that does the work between the apps

Arc is your personal AI assistant, built for the enterprise. You give it a goal, and it plans and coordinates the work needed to achieve it. It can take on tasks you assign or pick up work from the tools you already use, across apps, files, browsers, ServiceNow, and other enterprise systems. When something fails, it retries. It works with your own credentials, so it can access only what you can access, and it builds a memory of your work over time so you don't have to re-explain context.

Arc works where your work happens. Give it an outcome, and it plans the steps, moves across the systems involved, and completes the work, all within the policies your organization sets in Super Intelligence Control Tower. Workflows remain the foundation for the processes you've designed. Arc takes on the work between them, extending the ServiceNow AI Platform to the tasks no one has had time to automate.

What this looks like in practice

Here are a few examples of how teams might use ServiceNow Arc:

• Briefings before important conversations. Before a check-in, Arc can pull from your calendar, email, files, and ServiceNow records to build one briefing with context, prior notes, and recommended actions. You walk in informed, and your prep time goes where it matters most.
• 
• Reports on schedule. Arc can draft weekly SLA and backlog reports for service desk owners and stage the stakeholder update as an email for your review. Set it to run every Monday at 8:00 AM, and it keeps running without you.
• 
• Handovers and post-incident reviews. After an incident closes, Arc can draft the post-incident review from the incident history. At shift change, it can write the handover for the agent taking over the queue.
• 
• Adoption tracking. Arc can watch adoption across customer processes, and flag stalled processes and expansion signals, so Customer Success Managers spend less time on status calls and more on strategy.
• 

Governance you can set once and trust

Security and IT leaders should be able to say yes to Arc with confidence, so governance is built in at both ends. NVIDIA OpenShell secures the agent. Tool execution, connectors, and model calls run inside a secure sandbox packaged as a MicroVM, which limits the files, programs, and sites Arc can reach. ServiceNow SI Control Tower, which we call Super Intelligence Control Tower, governs Arc across the company.

Before any action runs, it passes through four gates, and any one of them can stop the work:

1. Permission: checks whether the tool is allowed. A block is a hard stop.
2. Approval: checks whether the action needs approval, and can pause for a human.
3. Containment: checks whether the action is contained by the sandbox and network rules.
4. Feature access: checks whether the feature is turned on for that user group.

All four have to say yes. Administrators set policy once, such as allowing users to read mail while requiring approval to send it, and users can tighten those limits for themselves. Destructive actions are hard-gated. Every model call is logged through the Generative AI Controller, and every agent action stays auditable in the SI Control Tower, which also provides a kill switch. Nothing runs unless the SI Control Tower policy allows it.

Model flexibility without the guesswork

Arc isn't tied to a single model family. It supports frontier model families like Google Gemini and Anthropic Claude (Opus) today, routed through the Generative AI Controller, and it's designed for per-task model routing, so each task can run on an appropriate model without you having to switch manually. Cost visibility by user and connector helps keep AI spend accountable.

Availability

ServiceNow Arc is generally available as of October 2026 on the BP1 release and is distributed through the ServiceNow Store. It's compatible with Australia Patch 7 and Zurich Patch 13. It runs on macOS today, with Windows support coming soon. Arc is licensed separately as part of the AI Workflow Factory SKU, and consumption is measured through Assist metering. Your ServiceNow account team can walk you through SKU composition, entitlements, and pricing.

Roles required:

• To access & use Arc:  sn_app_cowork.user 
• To configure its policies, set guadrials, observability, and audit: sn_app_cowork.admin

Getting started

Start by confirming your foundation with your account team: SI Control Tower for policy, observability, and audit; the Generative AI Controller for model routing and logging; and Action Fabric for access to platform data. NVIDIA OpenShell is packaged with the Arc installer, so there's nothing to deploy separately.

1. Admins install Arc from the ServiceNow Store.
2. Admins assign roles and connect the systems users need, such as Microsoft 365, GitHub, their ServiceNow instance, and other enterprise systems.
3. The AI Steward sets policy in the SI Control Tower.
4. Users install the Arc desktop app and sign in with their own credentials.
5. Users are live in minutes.

A good first step is to pick one recurring task that eats up time every week, like a sprint review, a stakeholder report, or pre-meeting prep. Have Arc do it once, refine the result, and then turn it into a scheduled routine.

Ready to see it in action?

Reach out to your ServiceNow account team to set up a briefing for your team, and stay tuned for more updates.

This article may contain forward-looking statements. Features and timing described are directional and subject to change.

Additional resources

• ServiceNow Arc Prod Docs
• ServiceNow Newsroom
• AI Workflow Factory Prime