Remember when “automation” was the big flex in telecom? If you were automating provisioning or ticketing, you were the cool kid at the standards table. Fast forward to 2025, and bragging about automation is like pulling out a BlackBerry at a 5G hackathon. Nostalgic, but not exactly where the future is headed.
The truth is automation did its job – it made processes faster, more repeatable, less error-prone. But there was always a need for it to be more. Scripts still needed fine-combing, workflows still broke at 2 a.m., and scaling seamlessly was extremely difficult.
Autonomy is the grown-up sibling. And it’s not just a word swap. It’s about building networks that can learn, adapt and heal themselves. Not because some consultant wrote a workflow, but because the network itself decides what’s best – and does it without asking permission.
Let’s dive a bit deeper.
Essentially, automation thrives in simple and complicated systems where behavior is predictable, repeatable, and rule-based. If you can write a playbook, automation can follow it. That’s why it worked well for provisioning, ticketing or routine fault management.
But telecom networks aren’t just complicated anymore; they’re complex. No rulebook truly applies. Today’s networks are a hairball of cloud, RAN, core, NTN, slicing, and private networks – all expected to deliver flawless QoS at real-time speed.
That’s why automation, which started with tasks, gradually expanded into processes, and is now shifting toward independent decision-making (autonomy). The next wave isn’t about telling machines what to do, but teaching them why – and letting them pick the best course of action.
That requires trust. For autonomy to succeed, transparency, explainability and reversibility aren’t “nice to haves.” They’re survival traits. If an AI agent changes a thousand parameters overnight, the CTO needs logs, reasoning and rollback. And accountability. Otherwise, you’re not running a network – you’re running a lottery.
Autonomy is the systematic layering of AI, SMO (Service Management & Orchestration), and apps (rApps/xApps/dApps) that abstract complexity while keeping the network open and intent-driven. It doesn’t run on rules; it runs on principles, policies, and guardrails. It operates through observation, feedback, adaptation, and continuous learning.
It doesn’t ask “What should I do if X happens?” It asks “What’s the best decision in this new, never-seen-before situation?”
Think of SMO as the Netflix of networks. In the past, launching a new service was like making a movie – big budget, multi-year production, closed distribution. With open SMO and rApps, anyone can launch innovation into the ecosystem. If it works, it scales. If it fails, it fails fast.
This shift also dissolves silos. Engineers don’t need to become coders; AI does the heavy lifting. An RF engineer would simply describe their troubleshooting steps in natural language, and an AI-powered platform would turn that into a reusable playbook. That’s autonomy in action. Here domain knowledge is amplified, not replaced.
But what about ownership and accountability?
Here’s the uncomfortable bit: technology is the easy part. Culture is hard.
Autonomy demands moving from “command and control” to “collaboration and trust.” It means ops teams that accept that closed-loop automation can act without human approval. It means governance models that balance speed with accountability. And it means reskilling: AI literacy for engineers, not just data scientists.
To quote my colleague Geoff Hollingworth, telco has a bad habit of chasing shiny objects – rebranding the same ideas every six months while startups die in the long procurement cycle. If we keep moving the goalposts, autonomy will remain a slide deck, not a reality.
Rakuten didn’t stumble into autonomy – it was designed in. From day zero, our networks were cloud-native, disaggregated and open. Autonomy isn’t a “future roadmap” bolted on top; it’s the logical outcome of the architecture we already run at scale.
Our approach boils down to three pillars:
And we’re clear on the business case: autonomy isn’t just cost optimization. It’s the only way to monetize at the speed of enterprises who demand customized, runtime-configured services – not static catalogs.
Automation was yesterday’s news. Autonomy is the present and future. But only if we recognize the fundamental shift:
Because the telcos who will ace the race to autonomy will not be the ones celebrating workflow scripts; they’re the ones building networks that never sleep, never silo and never wait for a manual approval to do the right thing.
To put it less politely: if you’re still running your network on “automation milestones” in 2025, you’re driving with cruise control in a world that’s moving to self-driving cars.
