How Agentic AI can make OT infrastructure more robust and secure
_{^{Leveraging telecom’s Dark NOC strategy to enhance your OT Network Operations Center resilience, efficiency and security}}

Imagine a sprawling oil refinery, where thousands of sensors, controllers, and automated systems orchestrate the transformation of crude oil into usable products. Every valve, pump, and pressure gauge is connected to a central network, feeding real-time data into control systems that ensure safety, efficiency, and compliance.

In this environment, disruption to the Operational Technology (OT) network can cascade with serious consequences, from halting production and incurring millions in losses to risking environmental damage and putting lives at risk.

Why it’s important to shed light on the Industrial NOC

Historically, IT and OT operated in silos: IT managed data, connectivity, and enterprise systems, while OT focused on process control, machinery, and safety. But with the rise of Industry 4.0 and digital transformation, we’re witnessing increasing convergence—a steady blurring of boundaries where OT data now flows into IT systems to power predictive maintenance, AI/ML optimization, business intelligence, and more.

This convergence addresses the need to enhance IT/OT integration and modernize Industrial Network Operations Centers—the infrastructure nerve centers where industrial systems (like MES, SCADA, Historians, and advanced analytics) operate and converge with IT systems. OT infrastructure availability is crucial for maintaining operational continuity and any disruption in these systems can lead to significant financial losses, physical damage to equipment, or severe societal impacts.

Below are a few examples of IT/OT convergence related network efforts increasing resilience:

• A Siemens plant in Texas faced frequent network-related production disruptions due to poor IT/OT integration. After redesigning their OT network, the company claims that downtime was reduced from 3,083 hours to just 15.4 hours and support calls dropped by 80%.
• Cisco deployed Software Defined Networking. According to Cisco, this enabled Audi to virtualize production assets and ensure deterministic, secure, and scalable networking across its manufacturing operations, resulting in increased manufacturing flexibility and network resilience.
• Faced with the challenge of modernizing its fragmented IT/OT infrastructure, a utility provider turned to Juniper’s zero-trust, programmable network fabric to unify operations from control centers to substations. Juniper claims that the transformation was so effective that SCADA circuit provisioning, which previously took hundreds of hours, was reduced to just minutes—dramatically improving efficiency and reliability.

The opportunity for next gen Industrial NOCs

Even with modernized infrastructure, network incidents continue to rely heavily on human monitoring and intervention. Typically, and for good reasons, the OT world runs on predefined alert thresholds and manual runbooks to manage problem solving. This ensures that fixes, adjustments, or restarts can be carried out without disrupting critical processes or causing serious consequences. The downside of this approach is reduced response time and agility during costly network downtime or periods of limited availability.

With the rise of Agentic AI technology now available, transformational benefits can be unlocked for industrial NOCs, such as real-time issue detection, analysis and proposed resolutions to improve uptime and scalable operations. Powered by strong human-AI chemistry, reliable agentic systems can balance autonomous capabilities with human-in-the-loop oversight. With such tools, people and operations can be augmented While Agentic AI is extremely ahead of the adoption curve, industrial organizations appear to have a strong appetite for deployment in industrial maintenance and support environments.

From the CRI Report “ER&D Trends”
71% of organizations believe AI will transform the maintenance and support activities (eg. AI-based diagnostics, predictive maintenance planning, intelligent support bots) over the next 2-3 years.

Agentic AI can significantly enhance Human-AI collaboration in Industrial Network Operations Centers by deploying intelligent agents that continuously monitor systems, detect anomalies, and perform root cause analysis in real time. These agents act as proactive partners, surfacing insights and recommended actions to human operators, enabling faster, more informed decision-making. By fostering a responsive and transparent interaction loop, Agentic AI builds trust and synergy between humans and machines to accelerate issue resolution, minimize downtime, and ensure operational resilience.

At the 2025 Mobile World Congress in Barcelona, we showcased a working proof of concept in collaboration with Exfo, Salesforce and ServiceNow, demonstrating our Agentic AI enhanced Dark NOC approach for telcos.

Can the Dark Industrial NOC concept work in the OT world?

The answer is ‘yes’, with the right approach. In the traditional context, the term “dark” in Network Operations Center (NOC) operations typically refers to a lights-out facility – one where no personnel are physically present because routine tasks are fully automated. The vision of a future-state Dark NOCs builds on this concept by introducing intelligent systems capable of observing, analyzing, and acting autonomously.

However, we believe that to be successful, you must go a step further, redefining the DARK concept from merely an automation milestone to a Digital Twin & Agentic AI-powered Reasoning Kernel (DARK) enabled NOC.

Our approach centers on a unified, AI-driven decision-making framework that enhances network management through deep contextual understanding and autonomous reasoning. At the core of this model are network digital twins, enriched with semantic models, which simulate and analyze network behavior with high fidelity. Importantly, this is not about eliminating human involvement; it’s about redefining the human-machine loop. In a DARK-enabled NOC, AI and automation handle routine detection, diagnostics, and remediation across domains. Human expertise is reserved for oversight, handling exceptions, and driving strategic evolution, not day-to-day firefighting.

A true Dark NOC is not a patchwork of scripts or rule-based automation. It requires systems that can understand context, correlate across silos, and reason to make decisions. Without semantically integrated data and agentic AI, this level of intelligence is unattainable.

Is it safe and robust?

Intelligent, self-operating network and operations centers in an industrial setting require a robust AI framework. This is why Capgemini developed the Resonance AI framework. It helps organizations unlock the full potential of AI at scale, and reinforces organizations’ AI readiness, while creating the right ‘human-AI chemistry’, to ensure long-lasting adoption.

Implications for decision makers: cost and revenue impact

For decision-makers, the short-term implications are clear: investing in advanced NOC capabilities must become a business imperative—not only for maintaining uptime, safety, and competitiveness, but also for unlocking cost and revenue potential, focusing on:

• Higher uptime for industrial networks can now be assured without scaling headcount, reducing operational costs while improving service continuity.
• A stronger security posture minimizes risk exposure and regulatory penalties, protecting both brand and bottom line.
• Enabling robotics-driven production environments—which demand superior network connectivity—removes a key constraint to achieving higher industrial efficiency and throughput.

Conclusion: Building the smart backbone of industrial innovation

As OT systems become more intelligent and interconnected, the networks that support them must evolve in tandem. The journey from complexity to autonomy in network operations is not just a technical challenge, it’s a business opportunity. Those who lead this transformation will define the future of industrial performance.