Research shows CSPs must break data barriers to succeed in delivering autonomous operations
While most communications service providers (87%) have started to implement AI into their network operations, either as proof of concepts or into production, only 57% have deployed telco AI use cases to the point of production.
Fewer than one in ten (6%) of CSP respondents surveyed in the research believe they are at zero-touch automation, which relies on AI and machine learning (ML) algorithms to manage and improve network operations.
The numbers were revealed in a Nokia-commissioned survey of 84 CSPs globally carried out by Analysys Mason.
One of the biggest hurdles to overcome is how to integrate AI into heterogeneous networks where access high-quality data sets is hampered by legacy systems with proprietary interfaces. Almost half of Tier-1 CSPs ranked data collection as the most challenging stage of the telco AI use case development cycle.
“CSPs must transition to more autonomous operations if they are to manage networks more efficiently and deliver on their main business priorities,” said Analysys Mason principal analyst Adaora Okeleke. “But as this research demonstrates, accessing high-quality data remains a critical obstacle to deploying telco AI within their networks. They need to really examine their AI implementation strategies to work around this data quality issue.”
Worse for telcos is that, according to McKinsey, AI investments are often not aligned with top-level management priorities; lacking that sponsorship, AI deployments stall, investment in technical talent withers and the technology remains immature.
Can see the value
The Nokia survey found CSP respondents said they believe AI will help improve network service quality, top-line growth, customer experience, and energy optimisation to meet their sustainability goals. As a result, there is plenty of proof-of-concept work happening and telcos are actively recruiting – from a small, highly-sought-after pool – AI specialists.
CSP respondents, primarily those in developed Asia–Pacific, North America and Western Europe are investing in AI to drive operational efficiencies for 5G networks. They need to grow revenue and maintain profitability to justify these investments, while also improving customer experience. Vodafone Italy, for example, implemented an anomaly detection solution using AI to automate network planning and optimisation functions.
Following this trial, Vodafone achieved increased operational efficiency of 25% to 30% because of the reduced time to detect and resolve issues and the associated cost savings gained from automating these workflows.
Japan-based Tier-1 CSP, KDDI, has proven in a recent trial that AI technologies can help to reduce energy utilisation at radio cells by up to 50%, without impairing customer experience. KDDI implemented a network energy management system that utilised ML to create models that analysed real-time demand and traffic patterns, and then automatically adjusted the amount of power consumed by RAN resources to match demand.
However, given some of the cost involved CSPs are not actively investing in AI platforms but will seek to use it as-a-service from AI cloud providers. This demonstrates why the new Global Telco AI alliance could prove so important.
Network use cases are appearing
The network use cases that have been most frequently deployed include network security, network design and planning use cases, with over 80% of respondents reporting that these use cases are in production. Customer care and experience use cases (such as AI-based chatbots, customer issue prediction and intelligent routing) have also been deployed in production by almost 60% of survey respondents.
About 20% of CSP respondents are investing in new services based on data insights derived from AI. These services include video analytics and IoT-related services such as smart manufacturing and autonomous driving.
Data siloes
With access to high quality data the number one issue for CSPs looking at AI, the research urges CSPs to evaluate their telco AI implementation strategies and develop a clear roadmap for AI implementation to overcome their data challenge and other impediments, such as an inability to scale AI use case deployments.
“A well-defined blueprint should help CSPs to develop a strategy to deploying AI use cases. This blueprint should consider which use cases to implement, and the timescales required to develop and deploy them,” the report suggests. “CSPs should start by identifying AI use cases that can address their top business priorities. This step is critical for ensuring that CSPs target value-generating use cases. Once use cases are selected, as well as the data sets to help them deploy the use cases, the relevant data sources (for example network equipment and OSS systems) should be identified.”
“CSPs are aware of the challenges of more deeply embedding AI into their operations and, as this research points out, the steps they can take to positively alter that situation, including building the right ecosystem of vendor partners with the right skillsets that can better cater to their network needs,” said Nokia cloud and network services head of business applications marketing Andrew Burrell.