Artificial intelligence (AI) for telcos created a huge buzz at this year’s MWC. It is being trialled or deployed by operators for many purposes from their own equivalents of the AI-driven digital assistants and chatbots of the web providers; to machine learning to make sense of all the data they hold on their subscribers’ usage and behaviors.
One of the most interesting areas is the use of AI to plan, manage and optimize networks, especially as part of the goal to reach a very high level of automation in the 5G network, to support a hyperdense platform with millions of moving parts and devices, which would be impossible to run manually.
Deutsche Telekom CEO Timotheus Hoettges was warning that telcos must retain the “human touch” when it comes to customer service, and not rely entirely on AI. “I don’t think we should try and copycat Amazon,” he said: “They will always be better at what they have started.”
However, when it comes to the network, DT has called for “brutal automation” to reduce costs and support large-scale IoT and densification programs, along with virtualization.
Nokia was the most advanced of the big OEMs in rising to that challenge last year, with a portfolio of services and tools to enable what it calls the “self-driving telco”. At MWC, Ericsson, Huawei and ZTE were catching up quickly. Ericsson was particularly active, unveiling “machine intelligence” products including a chatbot to support mobile network technicians, and a self-learning tool that automatically moves network resources between base stations as traffic patterns change. The latter is being trialled by Vodafone Spain, and if that goes well, it should be commercially deployed later this year.
Ulrika Jägare, director of analytics and machine learning, said the tools would also be used in Ericsson’s own managed services businesses. “It is better to have a self-learning and self-optimizing tool that adapts to each site than to have one model that is the same across the entire network,” she said.
Ericsson has also been exploring the use of machine intelligence in the core, with a virtual MME (mobility management entity) network function. “Compared with using a predefined product based on best assumptions, this is performing 25% better,” Jägare told LightReading. “You can easily reuse the AI for another product and we are looking at using it for the IMS function and in a virtual RAN.”
Huawei, in its pre-MWC announcements two weeks ago, promised an AI-enabled computing platform for telcos called Atlas, and it also offers Wireless Intelligence, a system for automating beamforming in 5G networks to improve efficiency and user experience. The vendor thinks 5G networks will have 50 times more configurable parameters than 4G.
Of course there is a downside to automation and AI. Telenor said recently that automation and digitization will mean the loss of 6,000 jobs by the end of 2020, about 20% of the workforce.
CEO Sigve Brekke said on the recent earnings call: “The load on the networks is not falling and data growth is exponential, and we are expected to service that and maintain a margin, and maybe even grow the margin. The only way to do that in the short term is to go heavily after opportunities such as digitizing the core and going for cloud-based solutions instead of lots of different data centers.”
The operator aims to virtualize 75% of its IT this year and to have 90% of network functions running in the cloud by 2020. A common network and IT platform should result in cost reduction of 20% to 30% by 2020, and an opex reduction of between 1% and 3% a year.
And Russia’s MTS has a similar message – it will cut about 1,000 jobs at its customer services unit this year as a result of automation and digitization, and expects, in future, a similar pattern to be seen in other departments such as network operations.