Telecom networks are more complex than ever. They include thousands of devices, multiple vendors, enterprise service-level agreements (SLAs), 5G slices running mission-critical services such as healthcare connectivity, and public safety considerations. The volume of data these networks generate has outpaced what teams can reasonably monitor and act on.
The traditional performance management model, which polls devices every five to 15 minutes, ingests data, runs the data through machine learning models to detect anomalies, and routes alerts to humans who decide whether to open tickets, is too slow and too brittle for today’s network demands.
When you're managing a network slice powering a hospital's connected devices, every 15 minutes isn’t good enough. What telecom operations need is AI that doesn't just surface a problem, but is designed to find it, understand it, and act on it.
This is exactly the kind of work that Project Arc, ServiceNow's new class of autonomous desktop AI agents, was designed for. It doesn't wait to be asked. It reasons, writes code, executes, and adapts when things don't go as expected, working continuously across enterprise tools and systems to complete complex, multistep work.
In a network operations center (NOC) context, that means an AI agent that can connect directly to network devices, run diagnostics, correlate what it finds against known thresholds and service dependencies, recommend remediation steps, and automatically create a ServiceNow incident and work order—all without a human in the loop until the moment one’s needed.
For a NOC operator, that's transformative. Instead of monitoring dashboards and triaging alerts, your team focuses on the issues that genuinely require human judgment. The routine, repetitive, high-volume work that accounts for most of the daily grind, including many tasks that are genuinely critical, now belongs to AI agents.
Here's where it gets interesting, and where a lot of autonomous AI deployments stumble. Giving an AI agent the ability to connect to live network devices, run commands, and create tickets in production systems is powerful. It's also exactly the kind of capability that keeps a chief information security officer (CISO) up at night.
ServiceNow and NVIDIA have built the trust layer designed to make this safe at enterprise scale. Every action Project Arc takes runs inside NVIDIA OpenShell, a secure, open-source runtime environment that sandboxes agent execution and enforces policy at the moment of action.
AI agents start with zero permissions. Every file read, command executed, and network connection is explicitly granted or blocked. Nothing gets through that hasn't been approved.
On top of that, ServiceNow AI Control Tower provides the enterprise governance layer. Every AI agent active across the fleet is discovered and monitored. Policies tied to user roles and system permissions are defined centrally and pushed to OpenShell at runtime. Security teams get full observability into every large language model (LLM) call and tool invocation and have the ability to terminate any agent instance immediately if something looks wrong.
Together, OpenShell and AI Control Tower mean autonomous is not synonymous with uncontrolled. You can intelligently delegate work within guardrails that your security team has authorized.
When a critical issue is detected—such as a power fault, a degraded optical interface, or an anomaly in a priority 5G slice—it doesn't just log it. It assesses severity, cross-references your network topology, recommends specific remediation steps with timelines, and asks if it should create an incident and work order.
If you say yes, it does so automatically, with a full audit trail.
If you're managing a network slice for a healthcare provider or a financial services firm with strict SLA commitments, you can configure an AI agent to run far more frequently than every 15 minutes. It can run every minute, if that's what the service demands. The cost of missing a fault in that environment far exceeds the cost of more frequent polling.
Because all of this executes through OpenShell and is governed by AI Control Tower, every bash command, device connection, and API call is logged, policy-checked, and auditable. The CISO has visibility. The workers in the NOC have relief. The network gets smarter.
The industry has been talking about autonomous networks for years. The gap between aspiration and reality has always come down to trust: Can we let AI act on the network without a human reviewing every step?
The answer is yes, but only when security and governance are built in from the start, not layered on afterward. ServiceNow AI Control Tower and NVIDIA OpenShell provide that foundation. Project Arc is the proof of concept demonstrating that the model works.
For telecom operators looking to move from reactive network operations to genuinely autonomous network management, this is where the journey starts.
Find out how ServiceNow can help you put AI to work for telecom.
