AI will expand rather than disrupt the RAN ecosystem

As the uncertainty and hype about AI’s exciting possibilities gradually resolve into solid real-world telecoms applications, a number of facts are emerging. First, AI offers an opportunity to fix historical mis-steps, such as creating success from faltering investments in self-organising networks (SONs). Second, AI will expand rather than disrupt the RAN ecosystem. Third, telecoms operators expect vendors to do the heavy lifting on AI. Finding the most appropriate applications for AI will be the product of careful planning and co-operation, rather than a headlong rush for change.

Transcript

Hello, I'm Caroline Gabriel. I lead all the research at Analysys Mason and have particular expertise in the operations and economics of mobile networks.

What are the main drivers for telecoms operators to incorporate AI into their mobile network operations ?

The main drivers to invest in AI for the mobile network were highlighted in a very recent survey that we conducted of 65 large operators companies around the world, and that was specifically about their reasons to be interested in investing. We saw a very high level of interest and the drivers for that were falling into two categories. One is efficiency and the other is quality of experience. So efficiency is largely relating to using AI intelligence to make better use of very valuable resources such as spectrum base station capacity and so on. And then, to build on that, to drive traffic management very intelligently and using the predictive power of AI to deliver the best experience to every site or every user. A secondary important driver is that a lot of operators have invested in first-generation network automation such as self-organising network, or SON, and seen quite disappointing results. So they're looking to AI to enhance the results of those investments, get a better result this time and, in effect, rescue the investment that they made a few years ago.

What risks does AI pose to telecoms operators and how should the value chain mitigate these risks?

The main risks that operators need to be aware of, again, were highlighted in the survey that we conducted and they fell into three main categories. One is that there is quite a shortage of historical data, particularly for 5G networks, which are quite new, to feed into the learning models. The second is that there aren't very many agreed industry-standard frameworks for the AI models and that sort of makes it more complicated for operators to implement their own AI strategies. And the third one relates to skills. There's a shortage of AI expertise, particularly in the telecoms sector. And also if operators are using AI for increased automation, they need to understand better the impact on their workforce. Will that reduce the workforce or just change the skill profile?

Is RAN AI entirely about efficiency or could it enable new revenue opportunities for telecoms operators?

We don't see RAN AI by itself significantly disrupting the RAN value chain. We do think that value chain will be expanded by AI; there'll be much more inclusion of companies or individuals who have deep expertise in AI, machine learning and big data. And many of those will come into the RAN sector for the first time, perhaps from IT, data or enterprise sectors. However, the RAN ecosystem is dominated by a few companies that have a very significant and long-standing base of knowledge in the fundamentals of operating around, such as traffic management. And we would expect those companies to remain in pole position and to include the AI experts into their own ecosystems. In fact, in our survey, about three-quarters of the respondents said that they would work on AI with their established vendors but would expect them to partner with experts from the data field. So we're not seeing significant disruption of the status quo, but we do see some expansion of that value chain to bring the skills that are needed into the operator market, particularly as most operators do not want to invest significantly in in-house AI skills of their own.

What is your one key message to the TMT industry right now?

For operators deploying AI in the RAN, our key message would be to start thinking about it now, but don't rush or don't get overtaken by the considerable hype that there is around the subject at the moment. This is very early stages in establishing the use cases and the business model and the ecosystem will mature over the coming years. So we would recommend that operators start to conduct trials and perhaps more importantly, consider the impact on their wider organisation before they start to implement very large projects.