Autonomous Network Transformation

This is about architecture — not a project timeline, not a migration schedule. It's about building an architecture for the next decade.

Back in 2014, Jio built the first-generation network management architecture — EPNM, CNAAP, and the tools around them. That architecture has been running for over 12 years now, all the way to 2026. It served its purpose remarkably well. But it was designed for a different era.

Now we need to do the same thing again — build the next-generation architecture that will serve Jio for the next 10 years. The migration and deployment itself will take 12 to 18 months. But the architecture we build during that period needs to last a decade. This is not about individual products — some products will evolve, some will get replaced, new technologies will get introduced. But the architecture itself will remain intact, as long as there's no highly disruptive technology shift in between. Just like the 2014 architecture lasted 12 years, this one is designed to do the same.

EPNM is already end-of-sale. End-of-support timelines are approaching. CNAAP follows a similar trajectory. What this means practically is — no new features, no architectural enhancements, and eventually no security patches or bug fixes. Staying on these platforms is not just a technology risk, it's an operational and security risk.

There's also a fundamental design philosophy gap. EPNM and CNAAP were designed for human operators. The entire experience is built around a UI — a dashboard where an engineer sits, clicks through screens, reads alarms, and manually decides what action to take. The system waits for a human. It doesn't think, it doesn't act, it doesn't learn. It's a tool — and it only works when someone is sitting in front of it.

In autonomous networks, the primary consumer of data is not a human — it's an AI agent. Agents don't need dashboards. They need APIs. They need structured data streams. They need to query, reason, and act in milliseconds — not wait for someone to click a button. The entire operating model flips from human-driven to machine-driven, with humans providing oversight rather than doing the work.

EPNM and CNAAP are monolithic applications — large, tightly coupled, single-process systems. Upgrading one component means upgrading everything. Scaling means buying bigger servers. A single failure can bring down the entire platform. They were designed in an era when software was shipped on DVDs and deployed once a year.

CNC and PCA are cloud-native from day one. They run as microservices on Kubernetes — each function is an independent container that can be scaled, updated, and healed independently. You can roll out a new feature without touching the rest of the platform. You can scale telemetry ingestion separately from fault processing. You can deploy across on-prem, hybrid, or public cloud. This is the agility that modern network operations demand.

And here's the deeper issue. These legacy monolithic platforms were never designed for AI. They lack the API-first architecture needed to feed real-time data to AI agents. They don't expose topology, telemetry, and fault data in the structured, programmable way that machine learning and agentic systems require. A monolithic system cannot stream events to Kafka, cannot serve data to vector databases, cannot respond to API calls from fifty AI agents running in parallel. You simply cannot build autonomous network operations on top of EPNM or CNAAP — the foundation doesn't support it.

This is why the migration to CNC, PCA, and HCO is so critical. CNC is built API-first from the ground up. Every piece of data — inventory, topology, faults, configuration — is accessible programmatically. PCA brings real-time telemetry correlation and predictive analytics. HCO adds multi-layer IP-optical correlation. Together, they create the programmable, data-rich foundation that AI agents need to operate.

Think of it this way — EPNM and CNAAP can take you to Autonomous Network Level 1, maybe Level 2 at best. You hit a ceiling. With CNC and PCA as the foundation, combined with the Crosswork AI Agent Framework, we can reach Level 4 autonomy, Level 4.5 in many use cases, and in some targeted scenarios, Level 5 — full autonomous operations with zero human intervention.

Summary: This is not about a project that will run for 10 years. The project runs for 12 to 18 months. But the architecture we build will serve Jio for the next decade — cloud-native, API-first, AI-ready, TM Forum aligned, and built on open standards. Just as the 2014 architecture carried Jio for 12 years, this architecture is designed to carry Jio into the autonomous network era and beyond. The question is not whether to migrate — it's how fast we can get there.