News
Concerns were raised around the quality of data and actual progress made on network automation
The reality of Artificial Intelligence (AI) and its potential to transform operators’ business models does not yet live up to the hype – and operators are not ready to embrace it.
This was the message delivered during a panel discussion at Total Telecom Congress, which saw Elisa, Telekom Austria, Hrvatski Telekom and Telenor Research came together to debate how to capitalise on the AI opportunity.
While all operators had started to look at AI – with Telenor Research even developing its own platform to sell to other operators – the panellists agreed that there was a long way to go.
Astrid Undheim, VP Analytics and AI at Telenor Research, said the problem with any new technology is distinguishing the hype from reality.
“One of the reasons AI has become popular is because of the huge data and the new powerful machine learning models,” she said. “I don’t believe we will capture value from these new technologies if we don’t embrace that. This means a new way of working which is much more software-based and much more development-based. What I struggle with, and I suppose the industry struggles with, is managing this in the vendor community as well. Every vendor is rebranding everything – anything that produces anything anywhere near data driven insights is being rebranded as AI. That is difficult to manage and a lot of managers might not understand the difference.”
Undheim had this viewpoint backed by Hrvoje Jerkovic, Director, Information Technology, at Hrvatski Telecom. Jerkovic agreed that the rebranding of solutions was certainly evident and said many telcos see AI as a huge risk.
“The rebranding from automation to AI doesn’t make any sense,” he said. “Most operators in Europe are still struggling with the automation – not because the technology is not available and the use cases are not good, but because this complete telco business is our bread and butter. It’s perceived as a huge risk to allow some Artificial Intelligence to monitor or steal our network. This is also one of the reasons from the operator side at least why things are not developing that fast in the AI direction as we are still on this automation level, a bit below the AI.”
Another concern around AI was raised by Mario Meir-Huber, Director AI at Telekom Austria, who highlighted that the data also needs to be improved before operators deploy AI solutions at scale.
“I hate the term AI as in 95 % of cases it is something that has just been renamed to AI,” said Meir-Huber. “The issue is, you can bring AI to your leadership team and your CEO can decide to install it, but it is not so easy as someone has to do the actual work. Most companies are struggling with data quality and having a big enough data base to apply AI to. If you apply an AI algorithm to the wrong data you would get a wrong network prediction, if we apply AI on the wrong data about a customer, we could have a lot of false positives and you could train the algorithm incorrectly. We need to invest a lot in data quality, solving the base, before we can go to the fancy stuff.”
