As AI in telecoms surges, operators are facing unprecedented traffic levels. Yet revenue is beginning to stall – forcing operators to rethink how networks are being built and run.

Autonomous networks – AI-driven systems that don’t require human intervention – have therefore been posited as a strategic solution to scale networks efficiently and to prioritise resilience. There are certainly benefits to the technology, with Capgemini research revealing in 2024 that operators were realising a 20% improvement in operational efficiency and an 18% reduction in network OpEx.

Capacity spoke with experts from Nokia, Ericsson and Ciena about the bold solutions autonomous networks can offer if telcos recognise their full potential.

Making a case for autonomy

As demand soars, operators are caught between a rock and a hard place, according to Ciena’s SVP of global marketing & communications, Rebecca Smith. She argued that, given traditional revenue streams are slowing, operators are no longer able to sustain next-generation, high-bandwidth services with legacy infrastructure.

“AI is now driving one of the most significant investment trends for the industry,” she told Capacity. The ultimate focus here isn’t just deploying technology for its own sake, but it’s about aggressively reducing operating costs, achieving tangible financial results and unlocking new revenue opportunities.”

If legacy operating models are no longer viable, operators are finding they must shift from manual management to outcome-driven, AI-enabled operations. The benefits of this include reduced operating costs, faster service delivery and the potential for new revenue opportunities in a sluggish market.

“Network autonomy offers personalised customer experiences, increased customer satisfaction by solving problems before they happen or opens new B2B opportunities via zero-touch slicing offerings,” explained Rodrigo Brito, VP of secure & autonomous networks at Nokia.

Telecommunications networks must also support industrial automation, mission-critical communications and AI workloads, which all have very different performance, resilience and security requirements, as Blessing Makumbe, head of cloud software and services at Ericsson North Europe, noted.

“Manual, script-driven operations simply cannot keep pace,” he said. “Autonomous networks are a strategic enabler for CSPs to unlock new business models, deliver differentiated connectivity and radically improve operational efficiency.”

Makumbe described this transition as a “paradigm shift” from manual, rules-based management to a more outcome-driven AI autonomy.

“Humans set business goals; the network executes and adapts in real time,” he added.

How the industry is responding to the automation era

Accelerating the shift towards autonomous networks has quickly become an important priority for network leaders across the connectivity industry.

To that end, organisations like the TM Forum have introduced a five-level autonomy mission that CSPs can use to measure their progress. Most CSPs currently sit around level two or three, with a handful having achieved level four.

“The first [to achieve level four status] was the Danish wholesale provider TDC NET in June 2025, alongside Ericsson,” Makumbe explained. “In an autonomous network, driven by intent, operators specify high-level business outcomes and the network continuously adjusts its behaviour to deliver them with minimal human intervention.”

Brito added: “Many telcos are taking serious steps to achieve level four by 2030. For example, stc Saudi Arabia are deploying Nokia’s AI-powered MantaRay AutoPilot for autonomous RAN operations with no human intervention. This has led to them achieving 15,000 autonomous corrective actions per hour.

He explained how Nokia delivers autonomous networks by combining its AI- and intent-based network operations, 360-degree observability, AI insights and closed-loop automation to “help customers predict issues, optimise performance and reduce manual operations”.

Although it’s hard to predict future growth, TM Forum survey with Bain & Company revealed that more than half of telecom leaders expect their organisations to reach level four of autonomous network operations within two years.

Likewise, Fortune Business Insights has predicted the global autonomous networks market to boom from US$8.56 billion in 2025 to more than $45 billion by 2034.

Smith explained how the next five years will look like a gradual, structured shift from “fragmented operations” towards “fully automated, data-driven networks”. She highlighted how Ciena is the only vendor to deliver open, modular AI-ready OSS through Blue Planet, multi-layer network control (Navigator Network Control Suite) and programmable infrastructure for autonomous networks.

“We combine the rich network telemetry generated by our infrastructure with operational intelligence from [these] platforms to create a trusted data foundation for advanced analytics and AI agents for various use cases,” she explained.

Looking ahead, telcos are starting to realise AI as a multi-year structural driver of network investment – with autonomous networking following the same long-term pathway.

“It is no longer optional, but it’s definitely not something you install over a weekend – it’s a real transformation,” Smith explained.

The rising tide of agentic AI – and what holds networks back

With technology industry giants creating their own models to turbocharge the agentic AI, our experts suggested there will be a gradual approach to fully autonomous networks.

First, initial agents will assist humans by making tasks simpler, including automating routine tasks and increasing productivity. Over time, as trust increases, network operations will become fully autonomous.

“The compounding effect on efficiency and service agility will become significant,” Makumbe explained. “The network evolution aligns with the broader AI trajectory.”

However, the industry still needs to confront significant challenges, which Brito said are currently around “data quality”, “semantics” and “governance”.

Data quality and semantics: Brito argued that AI models are “not reliable without high-quality data” and that data scientists are currently spending around “80% of their time on manual tasks to wrangle data”.

Governance: Given how essential networks are to the functioning of society, Brito said that AI agents “need guardrails and validation mechanisms” through things like “digital twins” and “human-in-the-loop validation” to “prevent them from making costly mistakes or going rogue”. He added that “future networks will have swarms of agents, and a policy framework is essential to manage their interactions, optimisation trade-offs and potential conflicts.”

He advocates for Nokia’s ability to combine centralised data management and governance with a “semantic ontology layer” to power the agents.

“Nokia Data Suite provides curated, multi-vendor data products that ensure quality and speed up the creation of new AI use cases and agents,” he said.

Smith added that Ciena customers are now actively taking steps and are “on the path to prepare their networks” for the shift towards automation.

“Autonomous networks – and specifically agentic AI – rely entirely on high-quality, real-time data,” she explained. “Without good data, AI cannot make reliable decisions.”

Ensuring supplied data is clean is a vital step to preparing AI agents safely, with Smith adding that CSPs have a responsibility to clean their network and services through large inventory, orchestration and assurance programmes.

“Lumen Technologies is currently working with Blue Planet to consolidate multiple legacy inventory systems to create a single, unified source of truth for their network assets,” Smith added. “By cleaning up their data and eliminating operational silos, Lumen is building a robust, AI-ready network foundation capable of supporting advanced AI agents.”

While networks are on an automation journey, the transformations are not ready at scale. Makumbe explained that an important step in the journey is to modernise towards 5G Standalone and 5G Advanced, which provides the necessary architecture to support the promise of agentic AI.

“Adopting a cloud operating model also allows networks to absorb rapid AI innovation without compromising reliability,” he added. “The foundations are being laid now through 5G Standalone and 5G Advanced, and Ericsson is evolving its portfolio accordingly to drive forward the industry’s vision for autonomous networks.”

Reaping the rewards of digital progress

As the industry continues to scale, autonomous networks are expected to radically transform agility, costs and the innovation capabilities of digital wholesalers and telcos alike. The benefits of such progress could include an enhanced customer experience, improved operational efficiency and increased network resilience.

“For wholesale providers, autonomous networking is about running the network faster, smarter and at scale,” Smith explained. “Increasing autonomy takes a lot of the manual effort out of service activation and day-to-day operations, which means new services can be turned up in minutes instead of weeks.”

In Ciena’s case, the company recently worked with a wholesale fibre provider whose manual OSS processes were holding back its service delivery. Smith explained that, by eliminating manual workflows and enabling true self-service provisioning, the company saw major benefits almost immediately.

“Their activation times dropped and closed-loop optimisation cut restoration times to under five minutes for critical services,” she explained. “Autonomous networking improves reliability by continuously monitoring the network, predicting issues before they escalate and automatically taking corrective action so customers see fewer disruptions and faster resolution when something does go wrong.”

As traffic continues to grow, autonomous networking can enable wholesale providers to scale capacity without also scaling complexity and headcount – creating a foundation for more flexible on-demand wholesale services. This is something customers expect more of.

Makumbe added that these benefits are already being seen in live deployments.

“Closed-loop automation is reducing manual interventions, truck rolls and energy consumption, while intent-driven management is letting leading CSPs and network operators express business goals and have the network automatically design, configure and assure the corresponding services, translating into shorter time to market for new wholesale propositions,” he explained.

Agentic AI and closed-loop automation form a critical part of Ericsson’s approach to autonomous networks, enabling networks to reason and act autonomously to deliver measurable business outcomes and trusted operations.

“Agentic AI represents a genuine change in how networks can be designed and operated,” Makumbe added. “The autonomous network is well-equipped to support mission-critical applications in public safety, healthcare, smart cities, autonomous transport and advanced manufacturing.”

He explained how Ericsson is currently developing multi-agent autonomous systems, including the recently introduced Agentic rApps as a Service on AWS that coordinates RAN optimisation workflows and Differentiated Support that brings on-site agentic AI into operations.

“These capabilities depend on deep observability and high-quality telemetry as a single source of truth,” he said. “Security cannot be an afterthought here. As agentic AI becomes more prevalent in critical national infrastructure, a zero-trust, security-by-design approach is non-negotiable.”

Mapping future growth

While it’s still too early to predict where long-term growth is headed, the future is promising. As AI continues to scale across the telecoms sector, those who are prepared for a dramatic ‘lifestyle change’ will certainly benefit from innovation opportunities. Those who are perhaps slower to adopt new technologies could face being left behind in the race. What’s clear, however, is that automation is happening, regardless.

According to a Nokia and TM Forum study, mobile RAN is a primary investment focus for automation, with 62% of respondents prioritising it due to its high-cost base and operational complexity.

Brito noted: “As the AI supercycle accelerates, with unpredictable traffic patterns and the need for real-time control, autonomous networks become essential to deliver adaptability, reliability and responsiveness.”

Fully autonomous networks will have the ability to anticipate business needs and solve problems before they arise. It will accelerate the most in areas where it can deliver the largest impact.

Smith explained: “The fastest growth will likely be tied to highly practical use cases – for example, deploying autonomous networks that can successfully undertake fault management, execute predictive maintenance and automatically adjust to changing traffic patterns without human intervention.”

This feature forms part of Capacity’s ITW edition of the magazine, which you can find online HERE.

Related stories

ITW 2026 is happening now: Here is your complete event guide

TikTok and T Wholesale on silent authentication: Rethinking secure login with network APIs

Neos Networks CEO: How the telecoms landscape will change in 2026

Amazon’s $9bn satellite deal that could crush Starlink; if Apple doesn’t blow it up first