How GenAI could change telecom

Barely a week goes by without a big Asian telco unveiling a new AI initiative. 

In the latest, China Telecom has set up a dedicated AI subsidiary, China Telecom AI Co., backed by 3 billion Chinese yuan (US$424 million) paid-in capital. Its scope will include AI software development, hardware sales and solutions.

The NTT Group has revealed it's been building out a generative AI service for corporate clients that it expects to launch next March. It's aiming for sales of more than 100 billion Japanese yen ($682 million) by 2027.

Korea's KT Corp has debuted a large language model (LLM) that allows enterprises to create AI services using their own data sets. Besides forging local partnerships, KT has begun selling the technology to southeast Asia, teaming up with a Thai IT firm to build a Thai-language LLM.

The most active telco is, without doubt, SK Telecom, which has developed its own Korean language-based chatbot and has invested $100 million in Silicon Valley startup Anthropic. It is working on adapting Anthropic's LLM Claude to telecom services and scenarios.

SKT describes its AI strategy as a 'pyramid' comprising chip firm and data center business at the base, AI transformation of existing services in the middle and development of new AI services at the top.

It's the headcount

It's far too early to get any sense of how these disparate efforts are going to fare in the market.

But a new report from Analysys Mason breaks down how telcos are deploying gen AI and how it might restructure the industry in the next dozen years.

Tom Rebbeck, head of Analysys Mason operator and IoT research, says that at the most basic level telcos are experimenting with gen AI for tasks such as meeting summaries, or are feeding transcripts from call centers to help generate suggestions for call agents.

But – surprise – the biggest direct impact will most likely be headcount.

During a podcast discussion, Rebbeck said gen AI would absorb some of the functions in virtually every part of a telecom operator, in particular areas such as customer service and sales.

He cited a trial in another industry where AI boosted call center productivity by 14% – which would be enormous if it could be replicated in telecom.

He said telco-specific functions such as network planning would also be affected by AI, but the impact would be limited.

"You still need lots of people to build the networks and maintain the networks and so on."

But this dynamic would drive retail telcos to become ultra-lean, run by just a tiny handful of staff.

"If you project that forward, you could have an extremely lean service provider with far fewer people than today, but the network side would still be relatively large," Rebbeck said.

This would further accelerate the split between retail and network that is already underway.

"So the impact could go well beyond reducing the number of people in the call center and to having an impact on the industry and the shape of the industry overall."