Autonomous networks - hype or opportunity?

Part of the challenge in separating fact from fiction lies in the very definition of autonomy. Much like autonomous cars, a telecoms network can become either partly or fully autonomous. TM Forum, a global industry association for service providers and suppliers in the telecoms industry, has devised a six-step scale of autonomy, against which progress on network automation can be measured. The maturity model ranges from 0 to 5, with 5 representing full autonomy. The most basic level of automation (level 1), describes a situation where a specific and repetitive task is automated to increase efficiency. A fully autonomous network (level 5) refers to full automation capabilities across multiple services and domains, and across the entire lifecycle of the network^[3].

Level 5 remains an elusive goal, promising closed-loop automation, in which a system is able to monitor and assess network conditions, traffic and resources in real-time, and use this feedback to adapt and optimise without the need for external intervention^[4]. This "zero-x" experience - promising zero wait, zero touch and zero trouble - would represent a fundamental shift in network management, away from "in the loop” systems managed through some human interaction, to an "on the loop" model where humans are solely responsible for oversight of autonomous systems^[5].

Unfortunately, current evidence suggests that operators are struggling to make their way up the automation maturity scale. According to a survey from the Capgemini Research Institute, around 84% of service providers remain at level 1 or 2 automation, with 60% aiming to reach level 3 by 2028^[6]. Level 3 is described as a “conditional” AN that can sense real-time environmental changes, and adjust and optimise itself to these external changes within certain domains. Getting to stage 3 alone requires extensive planning and preparation - with only one in five of those surveyed having devised a comprehensive strategy to reach level 3 automation.

Few survey respondents expect to go further - just 16% of those surveyed are looking to reach level 4 autonomy, with 1% of survey respondents aiming for full level 5. These muted expectations are indicative of the difficulties in going beyond level 3 automation. Indeed, level 4 is a much more proactive and predictive system that incorporates deeper cross-domain integration than lower levels. Level 5 requires even deeper integration, embedding AI and ML-native operations throughout every aspect of the company.

Reaching levels 4 and 5 in the maturity cycle is not an impossible task, but there are several obstacles in the way. One of the most prominent issues, all too common to the telecoms sector, is fragmentation. Many in the sector subscribe to the notion that the trajectory from 0 to 5 on the scale will occur in stages, the first focused on leveraging the benefits of automation for highly specific, functional and priority use cases, before later scaling these solutions to fit different operational scenarios. However, by taking the more straightforward step of prioritising use cases first, operators can often end up devising varying business solutions for different customers and under different regulatory guidelines, making them hard to scale. Devising more open, case-agnostic and functional solutions can help overcome this problem and improve scalability.

Fragmentation also extends to the pattern of historical investment in network infrastructure, with many operators managing a patchwork of legacy systems, protocols and interfaces, whose varying composition can be hard to bridge. Very few operators have built up a “greenfield” network, built from scratch and absent of any constraints or limitations. In that sense, operators need to tread carefully and improve their legacy integration capability, avoiding the pitfall of creating multiple “islands” of automation across different domains and environments. Better cross-domain orchestration can ensure processes and domains are seamlessly integrated and connected, and capable of absorbing any future system alterations or wholesale replacements^[7].

Ensuring interoperability between these legacy systems and new autonomous networks is a complex and costly undertaking that involves significant capital expenditure and employee training^[8]. The challenge is often further exacerbated by the fragmented data landscape many operators face. Shortfalls in the quality and integrity of structured and unstructured data obtained from a variety of sources can make it difficult to derive insights and assist with decision-making.

Cybersecurity concerns also loom large. Privacy and security have long been regarded as critical considerations for the telecoms sector - these concerns take on even greater significance with autonomous networks that will be expected to execute decisions independent of human control or intervention. Given potentially sensitive user information can be held on autonomous systems, privacy controls, encryption and secure data management protocols must be put in place to maintain privacy and instil trust^[9]. If not managed correctly, ANs could potentially violate data protection regulation or broaden the digital attack surface, making networks more vulnerable to cyber threats.