Anand Naidu brings a wealth of expertise to the table as a seasoned development expert, specializing in both frontend and backend architectures. With a deep mastery of various coding languages, he has spent years dissecting the complexities of telecommunications infrastructure and cloud-native integration. His perspective is particularly valuable as we move into a new era where artificial intelligence does more than just analyze data—it acts autonomously to manage the global networks we rely on every day.
The telecommunications industry is currently undergoing a massive transformation, shifting from traditional manual operations toward an “agentic” model. This transition involves embedding intelligence across the entire stack to create self-healing networks that can predict failures before they happen. Throughout this interview, we explore the rise of autonomous agents, the technical hurdles of building digital twins with GraphML, and how carriers are leveraging these advancements to drastically reduce operational expenses while making the underlying network infrastructure essentially invisible to the end user.
The industry is moving toward an “agentic” model where AI handles reasoning and cross-functional workflows autonomously. How do these agents prioritize competing tasks during peak network congestion, and what specific steps are necessary to ensure their reasoning aligns with complex business logic?
Prioritization in an agentic model is driven by what we call business intent, where the AI doesn’t just look at signal strength but evaluates the commercial impact of its actions. When congestion peaks, these agents utilize an agent-to-agent protocol to communicate across the platform, allowing them to weigh a technical fix against the specific subscriber needs, such as ensuring high-value enterprise traffic remains stable. To align this with business logic, we must embed the intelligence directly into the BSS and OSS layers, creating a framework where a network event is automatically translated into a commercial outcome. This requires a prescriptive architectural approach, similar to how we manage our own two million miles of fiber, ensuring that every autonomous decision supports the broader goals of reducing churn and maintaining strict service levels. By orchestrating this single thread of intelligence, the agent can initiate a trouble ticket and dispatch a field technician while simultaneously notifying the contact center to inform affected customers.
Reducing network management event times by over 90% or cutting customer complaints by a quarter requires a fundamental shift in operations. Can you share an anecdote involving a specific network failure where an autonomous agent outperformed a human team, and what metrics best track this transition?
We have seen incredible results with partners like Deutsche Telekom, where the implementation of the RAN Guardian agent reduced the average time of network management events by a staggering 95%. In a typical manual scenario, identifying a radio access network anomaly, diagnosing the root cause, and coordinating a fix could take hours or even days, but the autonomous agent can detect and remediate these issues almost instantaneously. Another powerful example is Bell Canada, which utilized these intelligent workflows to reduce customer complaints by 25% by proactively addressing service degradations before the user even noticed. The primary metrics we use to track this shift are Mean Time to Repair (MTTR), the rate of proactive versus reactive trouble tickets, and overall customer churn figures. These numbers prove that moving away from human-led manual intervention toward an autonomous framework creates real, measurable value for the carrier and a frictionless experience for the subscriber.
Integrating intelligence across the full stack aims to make the underlying network infrastructure invisible to the end user. How does this integration change the daily workflow for field technicians and contact center staff, and what are the practical implications for maintaining service levels during this shift?
The daily workflow shifts from a “search and find” mission to a “verify and execute” model, where field technicians are no longer wandering through data to find a fault but are dispatched by an agent that has already diagnosed the problem. When the system detects an anomaly, it can autonomously trigger a dispatch and provide the technician with the exact coordinates and nature of the failure, making their job significantly more efficient. For contact center staff, the change is equally dramatic because they receive real-time updates on which specific subscribers are affected, allowing them to be proactive rather than reactive. Instead of taking a call from a frustrated customer, the center can send a notification before the customer even picks up the phone, which is essential for maintaining the “human-like” and tailored experience consumers now demand. This integration ensures that service levels remain high because the intelligence is embedded across the full stack, bridging the gap between a technical glitch and the customer support team.
Using GraphML to mathematically model how a fiber cut propagates through a digital twin provides a new level of predictability. What are the primary technical hurdles in building these high-fidelity digital twins, and how do they change the way carriers approach proactive maintenance and resource allocation?
The biggest technical hurdle is the sheer complexity of mapping thousands of interconnected nodes and legacy systems into a unified Graph Neural Network that can accurately simulate real-world physics. In our pilot project with MasOrange and NetAI, we focus on creating a digital twin that can mathematically model exactly how a single fiber cut will ripple through the rest of the network architecture. This level of fidelity allows carriers to move away from reactive “firefighting” and toward a model of proactive maintenance where resources are allocated based on predicted failures. By understanding the propagation of an incident before it happens, a carrier can reroute traffic or stage repair crews in high-risk areas, turning connectivity from a simple utility into a highly resilient, value-creating engine. This transition requires unlocking data that was previously siloed in disparate legacy systems, which is a significant undertaking but necessary for true predictive autonomy.
Scaling autonomous remediation from specific radio access networks to a global, network-wide framework presents significant deployment challenges. When implementing these intelligent agents across diverse international properties, how do you handle localized regulatory constraints while maintaining a single, unified thread of intelligence?
Expanding a tool like the RAN Guardian into a network-wide capability, such as the “Minder” agent we are deploying across all Deutsche Telekom properties, requires a very modular approach to AI. We manage localized regulatory constraints by ensuring that while the “reasoning” layer of the agent remains consistent, the “action” layer can be tuned to follow specific regional policies regarding data privacy and spectrum usage. This allows us to maintain a single thread of intelligence across diverse international properties while respecting the unique legal environments of each country. For example, our work with Vodafone involves leveraging an autonomous network framework globally, but the business intent and remediation steps are tailored to the specific operational needs of each local market. This balance is what allows us to scale efficiently without sacrificing the granular control required for telecommunications compliance.
Connecting a technical network event directly to a commercial business outcome requires embedding AI within legacy BSS and OSS systems. Which specific datasets must be unlocked to bridge this gap, and what are the long-term trade-offs between reducing operational expenditures and investing in these agentic platforms?
To bridge the gap between the network and the bank account, we must unlock and integrate subscriber usage patterns, real-time billing data, and historical churn indicators with the raw telemetry from the network hardware. By embedding Gemini Enterprise into these legacy BSS and OSS systems, we can see how a specific cell tower outage impacts the lifetime value of the customers connected to it. The long-term trade-off is often a higher initial investment in data modernization and AI integration, but this is offset by a massive reduction in operational expenditure (OpEx) as manual workflows are replaced by autonomous agents. Ultimately, the goal is to stop treating the network as a separate entity from the business and instead view it as an integrated platform where every technical action is a business decision. Carriers that make this investment now are the ones who will successfully transition from being “bit pipes” to becoming truly “agentic” enterprises by 2026.
What is your forecast for the agentic telco?
I believe that by 2026, we will see the emergence of the truly invisible network, where the “agentic telco” becomes the standard rather than the exception. We are already seeing the groundwork laid with our partners who are achieving 95% reductions in event times, and as these agents become more sophisticated, they will handle the majority of network operations without any human intervention at all. The future will be defined by “one Google in every hand,” where the complexity of the infrastructure is completely hidden from the user, replaced by a seamless, frictionless experience that feels instantaneous. We will move beyond just fixing problems to a state of continuous intelligence, where the network evolves and optimizes itself in real-time based on the commercial and personal needs of every individual subscriber.
