Photo by Akil Mazumder on Pexels

The Autonomy Gap: Telcos Are Further Behind Than They Think

The vision of a fully self-healing, self-optimizing telecommunications network has captivated the industry for the better part of a decade. Standards bodies like the TM Forum have built entire frameworks — most notably the Autonomous Networks framework with its six-level maturity scale — to chart the industry’s progress toward that goal. Yet for all the conference keynotes, vendor promises, and R&D expenditure, the uncomfortable reality is that the vast majority of telecom operators today sit at Level 1 or Level 2 on that scale: largely manual operations with only isolated pockets of rule-based automation.

The question worth asking is why. In an era when artificial intelligence is rewriting entire industries seemingly overnight, why is the telecom sector — one of the most data-rich, infrastructure-intensive industries on the planet — struggling to cross the threshold into genuine operational autonomy?

It’s Not a Technology Problem

The instinctive answer is to point at technology. Operators need better AI models, more compute, smarter algorithms. But that framing misses the deeper issue entirely. The tools required for autonomous network operations are largely available today. What’s holding operators back is the foundational layer beneath the technology: data quality, systems integration, and governance.

Consider data quality first. Autonomous network functions — whether predictive fault detection, dynamic spectrum allocation, or self-optimizing radio access networks — are only as reliable as the data fed into them. Most large operators have accumulated decades of network data stored across siloed operational support systems (OSS) and business support systems (BSS), often in incompatible formats, with inconsistent labeling and significant gaps in historical records. Training an AI model on dirty, fragmented data doesn’t produce a smarter network — it produces a confidently wrong one.

Integration readiness compounds the problem. A telco’s operational environment typically includes hundreds of network elements from multiple vendors, each with proprietary interfaces and management systems. Even when individual domains — the radio access network, the core, the transport layer — begin to develop autonomous capabilities, making those capabilities work in concert requires deep integration that most operators haven’t yet achieved. Autonomy at the domain level doesn’t automatically translate to autonomy at the network level.

Why Maturity Varies Wildly by Operational Domain

One of the most telling characteristics of the current autonomy landscape is how unevenly progress is distributed — not just between operators, but between different operational domains within the same operator.

The radio access network (RAN) is generally the most advanced domain when it comes to automation. Self-Organizing Network (SON) capabilities have been embedded in mobile infrastructure for years, and the emergence of Open RAN architectures is accelerating the integration of AI-driven RAN Intelligent Controllers (RICs). In contrast, core network automation — particularly around service assurance and end-to-end orchestration — lags considerably behind, weighed down by legacy virtualization platforms and complex interdependencies.

Customer-facing operations present yet another profile. Automated fault resolution that affects subscribers triggers regulatory, reputational, and contractual considerations that don’t apply to pure infrastructure optimization. This is where governance frameworks become critical. Without clear policies defining what decisions an autonomous system is empowered to make, operators default to human oversight — which means the automation stops exactly at the point where it would deliver the most value.

The Governance Deficit

Governance is arguably the most underappreciated barrier to autonomous network maturity. Operators have invested heavily in AI tooling but have been far slower to develop the organizational structures, accountability frameworks, and regulatory alignment needed to let those tools actually operate autonomously. Who is responsible when an AI-driven decision causes a service outage? How do you audit an autonomous system’s choices for compliance with network neutrality regulations? These aren’t abstract questions — they are active blockers on the path from Level 2 to Level 3 and beyond.

The TM Forum’s Autonomous Networks Reference Architecture explicitly calls for intent-based management, where business objectives are translated into network behavior without human micromanagement. But intent-based systems require operators to have a clear, codified understanding of their own operational priorities — a deceptively complex organizational challenge that many carriers have yet to tackle seriously.

The Path Forward: Foundation Before Ambition

For operators serious about advancing their autonomy maturity, the priority list looks less like a technology roadmap and more like a data engineering and organizational transformation project. That means investing in unified data platforms that normalize and contextualize network telemetry across domains. It means building model-driven interfaces — leveraging standards like YANG and NETCONF — to create the integration fabric that autonomous systems require. And it means developing internal governance structures that define the decision boundaries for AI systems before those systems are deployed.

Vendors and system integrators also have a role to play, shifting from selling autonomous network capabilities as standalone products to delivering them as part of broader operational transformation engagements that address the data and governance foundations from the ground up.

Industry Outlook

The TM Forum projects that fully autonomous networks — Level 4 and above — will begin to emerge in limited deployments by the late 2020s, primarily among Tier 1 operators with the scale and resources to make the foundational investments. For the broader industry, however, meaningful progress will depend on a shift in mindset: recognizing that the journey to autonomous networks is not a software upgrade but a multi-year operational transformation. The operators that begin laying the data, integration, and governance groundwork today will be the ones who find themselves capable of genuine autonomy when the technology matures enough to demand it.