Rohan Lobo, Partner, Consulting, Deloitte Touche Tohmatsu India LLP

Communication service providers (CSPs) are beginning to look at themselves as digital service providers. They are adopting a “platform approach” to provision and monetise services that are flexible and resemble OTT services. Perhaps, the most important change is that CSPs are replacing the functional view of customer service with an agile end-to-end view.

Central to this transformation is the use of automation, primarily to make processes more efficient. Machine learning (ML) and artificial intelligence (AI) are already being leveraged to predict customer behaviour, improve the hiring process, automate IT infrastructure, manage contracts and detect cyberthreats. Apart from this, interesting use cases are being observed in the network domain, that is, network services provisioning, network services architecture, and network engineering and operations. Technologies such as service orchestration, network function virtualisation and automation solutions such as self-organising networks (SON) in radio access are expected to reduce the repetitive network maintenance work by 70-90 per cent and reduce more than a third of skilled resource (manpower) requirements when it comes to RAN management. Some of the key automation and AI focus areas for CSPs are:

Automated optimisation

IP-based networks require the optimisation of several configurations and parameters in various components across the core, transmission and access layers of the network. Real-time subscriber event data at the individual and cluster levels could be analysed by machine learning algorithms to make site-specific changes such as power and antenna tilts (e-tilts), which could dramatically improve the quality of service and reduce operating costs. Similarly, dynamic carrier configurations at the site level could be determined based on usage to optimise energy consumption at the site, thereby reducing the opex.

Configuration and policy automation

CSPs have complex architectures, often operating multiple technologies in parallel (2G, 3G, 4G and 5G). In this context, the seamless configuration of various RAN platforms becomes increasingly important. Most CSPs now have orchestration layers that support the automation of policy control and enforcement, thus ensuring consistent deployment of parameters.

Zero-touch roll-out and provisioning

Pre-optimised configuration scripts for automated network design and integration increase efficiency and capacity to roll out more sites. ML also establishes a correlation between traffic trends and site usage, and helps forecast the impact of traffic management, thereby improving asset usage. Optimised offloading to nearby assets increases capex efficiency and reduces overall network costs. Wit the help of service, configuration and performance data, it is also possible to detect, diagnose and forecast events and trends. These diagnostics can be real-time triggers of optimisation actions implemented by RAN orchestrators and can enable increasingly fast and adaptable networks while minimising human errors.

NOC automation

The use of ML to correlate alarms in network operation centres (NOCs) and generate meaningful insights can improve troubleshooting accuracy and responsiveness, and ensure automated change of configurations. Predictive network maintenance also reduces site visits of field service engineers, thereby reducing opex.


Automating the process of executing handoff and compensation techniques to seamlessly offload workloads to healthy surrounding nodes before initiating the troubleshooting procedure can have a significant impact on user experience. Cost reduction is also a product of automation as these are technically challenging activities and often lead to long localised service shutdowns.


Most CSPs are grappling with potential models to scale up deployments – buy, build or partner to create solutions. The challenge of integrating AI into existing businesses may be amplified by fears of employees and resistance to the changes brought about by these technologies. Further, accurate, high quality data is the fuel that powers AI systems. Protecting the security and privacy of data presents another critical challenge. However, CSPs that are constrained by capex and opex are prioritising investments to simplify operations and improve service flexibility. The potential data struggles for organisations include locating and accessing the right internal and external data sources, cleaning and aggregating data from disparate systems, and ensuring that it is free from bias.

CSPs that have in-house experience to assess technologies that solve cross-functional problems will be able to better identify and implement solutions. With network quality and service flexibility being the prime drivers of the telco market, applications that integrate end-to-end cross-functional processes will be key to achieving visible and sustainable results.