AI in telecoms: practical advice and next-steps for carriers deploying AI
The hype surrounding the potentially transformative impact of artificial intelligence (AI) and machine learning (ML) on the telecoms industry is impossible to ignore. AI was the talk of the town at Mobile World Congress in Barcelona, again, at the end of last month, as a means to impose order on new 5G and IoT networks. It is riding high on Gartner’s hype-cycle of emerging tech trends, as well.
But progress is rather slow, so far, generally restricted to automating actions in carriers’ care channels and marketing functions. Here, AI experts from the telecoms industry advise on how the carrier community can take it beyond the hype, to deploy practical AI and ML applications in all aspects of network operations.
Francisco José Montalvo, director of ‘fourth platform’, Telefónica:
“It may sound cliché but, as with other changes in our sector, implementing AI and ML is not just a question of technology, people, and skills. It starts with a cultural change at all levels of the organization – trusting that AI can be complicated to start with, and that managing the cultural change is essential to accelerate its adoption across a business. Network operators can manage this cultural change in several ways, including with the launch of AI/ML culture programs, education and training with external experts, and redesigning processes so they can harness the power of AI.”
Andrew Burrell, head of marketing for ultra broadband and analytics services, Nokia:
“Our approach is driven by use cases and how we solve operators’ problems, or make their customers lives easier. So the technology discussion about AI is inclusive but certainly not the starting point. While customers sometimes ask how AI impacts delivery – does it require special infrastructure, for example – the answer is it’s embedded in software and doesn’t significantly change the infrastructure requirements. The other question that comes sometimes is a request to have a deeper understanding of the algorithms, but this usually comes further into the discussions about what we have that addresses their area of need.
“There is sometimes scepticism, and even fear about the use of AI. Successful adoption requires clear communications to establish the goals, and effective change management to implement new processes and measure progress towards those goals. Adoption is more successful if the users feel they have a clear understanding of these goals, and visibility on what AI is doing. So far open-loop and semi-supervised models have been used, and the evolution towards fully autonomous AI needs to follow a stepwise approach.”
Ben Azvine, head of security futures, BT:
“ML is based on data, so network operators should have a coherent approach to collecting and securing their data before they can use ML at scale. Some level of skill in AI/ML is also essential as there is a lot of hype about what is real in AI, so either having a small group of experts internally or partnering with reputable companies with experience in implementing AI is also valuable in understanding ‘dos-and don’ts’. Finally, it is important to raise the awareness of AI opportunities and challenges across the company through company-wide skills development initiatives.”
Zhang Sihong, chief engineer of AI solutions, ZTE:
“Implementing AI and ML should be considered from two aspects. The first is around network intelligence – operators should be demand-oriented, invest in AI projects, work with the industry, participate in the standards and open-source organizations, and study the algorithm. This will greatly speed up the level of network intelligence. The other point is around the intelligent customer experience – operators can actively strengthen the intelligence of customers’ equipment by introducing products.”
Jay Perrett, founder and chief technology officer, Aria Networks:
“It can be risky to rush ahead with a direct implementation of AI and ML as technologies, absent of a clear business problem to solve. What matters first and foremost is a clear idea of where a business process has a bottleneck; an understanding of why that problem exists, and whether AI could be used to addressit. Particularly in the area of networks, AI and ML alone are not enough. A chart or a dashboard generated by an AI platform does not constitute a network design. Our advice would be to start with a business problem that needs solving first, and see whether the nature of that problem lends itself to an AI or ML technology set.
“Another factor to consider is the willingness of the people affected to accept the results from an AI system. There can be tremendous emotion or political attachment to decades of received wisdom or convention. Embracing AI and ML certainly represents a serious cultural change, so it’s important to consider that. Without that, initiatives are likely to become stuck in research mode, and never make it out of labs or innovation hubs.”
Dmitry Kurbatov, head of telecommunications security, Positive Technologies
“They have to keep calm and carry on. It is always the same – the need to develop and maintain a clear strategy, constant development, and optimization, and to keep searching for new opportunities. And of course, one should never forget about security. It is easy to get carried away with the innovation but it’s crucial to remember you’re not the only one on the edge of technology – hackers are innovating too.”
Click here to register for the Enterprise IoT Insights webinar on March 21st on the developing role of AI in telecoms. Look out for the editorial report, ‘Artificial intelligence and machine learning: making IoT work for telecoms’, also published March 21st.