Can telcos sell AI agents? SK Telecom will find outCan telcos sell AI agents? SK Telecom will find out

South Korea's SK Telecom is preparing to launch its first major consumer AI service, Aster. The offering will test the waters for AI services provided by telecom network operators.

Mike Dano, Editorial Director, 5G & Mobile Strategies

January 22, 2025

7 Min Read
Futuristic robot at the MWC Barcelona trade show
(Source: GSMA)

SK Telecom made an interesting announcement at the recent CES trade show: It's developing an AI app called Aster that it plans to test in the US starting in March.

Why would a South Korean telecom network operator first release new AI tech in the US, I wondered? The answer helps highlight the Wild West nature of the AI industry, the role of the US in the global AI marketplace, the apparent desperation of some 5G players and the weird future we're all hurtling toward.

After exchanging some messages with SK's public relations team, I set up an interview with Jorge Clark-Tan, who's leading US marketing for the Aster initiative. He's the head of marketing for Global AI Platform (GAP) Corp., a startup that's fully funded by SK Telecom. Interested users can sign up for the beta service at www.asterapp.ai.

SK's GAP plans a full commercial launch of Aster in the latter half of 2025 followed by a global rollout in 2026. It'll start as an iOS app and expand to other platforms (like Android and the web) later.

At the beginning of our call, Clark-Tan explained that he couldn't speak for SK, only GAP. He also was very careful in the way he answered my questions in order to adhere to SK's talking points on AI. This cautious approach to public communications is pretty common across Asian telecom companies – it's basically the exact opposite of the sometimes painful honesty that many European telecom executives take with their public comments. US telecom officials often take a path between these two extremes. (These are the fun cultural differences you get to learn while covering the global telecom industry.)

Clark-Tan offered a few reasons why SK and GAP plan to release Aster in the US first and not in SK's home country of South Korea. First, he said, the US market represents the "gold standard" in product marketing. If you can get it right in the US, you can right anywhere, he explained.

But that's probably not the main reason. Clark-Tan also said it will be easier to develop an AI product in English in the US and then export it into South Korea, rather than the other way around. But perhaps most importantly, he said the US is a world leader in developing AI technology – including the English-based large language models (LLMs) underpinning most AI operations – making it ground zero for most AI services.

Welcome to the US, Aster

But what exactly is Aster and how will it compete with advanced AI services like ChatGPT, Gemini and Grok that are already being offered for free in the US?

Clark-Tan told me that Aster isn't a new LLM. Instead, it's built on top of some of the market's best LLMs (like Perpelexity and Anthropic), and it will use whichever LLM provides the best answers. But the real difference with Aster is that it will offer advice that's specifically customized to individual users.

Aster will learn about a particular user's likes and dislikes and provide recommendations accordingly, Clark-Tan explained. For example, Aster might know about a user's shellfish allergy and therefore filter out seafood restaurants from a list of where to eat in Las Vegas during CES.

"You kind of get the benefit of multiple LLMs," Clark-Tan told me.

That layer of customization might not sound like much to casual AI users (like me). But for others, it might make a huge difference. See: this crazy New York Times story about a 28-year-old woman falling in love with a ChatGPT boyfriend, who undergoes a personality lobotomy every few weeks because that's how long ChatGPT stores user preferences.

Targeting other telcos

But it's SK's approach to the AI market that is perhaps most interesting. Clark-Tan said the company plans to spend this year dressing Aster up for its commercial debut in order to then begin selling the offering as a white-label service to other companies, including other telecom operators.

Meaning, a US-based MVNO would be able to create a self-branded version of Aster and offer it to customers using the data it already has about those customers. Such an offering – a smart and helpful AI buddy from your carrier that already knows stuff about you – might go a long way toward improving customer satisfaction. It would at least provide a telecom answer to smartphone AI services like Apple Intelligence.

It would also be the latest use of AI in telecom. AI dabbling in the industry already stretches across customer service, internal operations and network management, just to name a few areas.

Into the great unknown

Aster, of course, faces an uncertain future. Already behemoths like Google and Microsoft are investing billions of dollars in the AI marketplace. It's not clear what space might remain for other, smaller AI providers.

Moreover, the age of "agent AIs" is not quite here yet. Most AI services can answer questions, but few can reliably perform autonomous tasks on behalf of a user. For example, after making its own autonomous decision, a true AI agent could then move a robotic arm, generate new content, or take any other physical or digital action – and then learn from that action.

But that kind of service seems to be coming, and soon. According to TechCrunch, OpenAI may be close to releasing an AI agent that can take control of a user's PC and perform actions on their behalf.

That's the kind of "AI buddy" service that Aster and others are working toward.

Another big AI question is whether power and telecom providers will be able to handle widespread use of services like Aster.

On the telecom side, companies like Lumen Technologies and Lightpath are already boasting of billion-dollar AI data transport deals with hyperscale web companies. But it's not clear whether 5G network operators in the US and elsewhere will be able to supply latency connections quick enough for end users to have real-time AI conversations. Right now, 5G latency speeds of around 30-40 milliseconds are still mostly in "hang on while I look that up for you" territory.

On the power side, things are so tight that President Trump name-checked the "next generation of technology" in new executive orders intended to expand US energy production. But there is no quick fix for rising power demands among AI data center operators.

The peak of inflated expectations

More broadly Aster is a great example of how telecom companies are quickly pivoting to AI as a way to inflate their fortunes.

The shift is driven in part by the failure of 5G to live up to its early, sky-high expectations. 5G was promoted by some as a way to usher in the world's "fourth industrial revolution." Instead it raised some users' speeds and not much else, including operators' financials.

As it became clear that 5G wasn't going to revolutionize much of anything, operators across the world began pivoting to the next big thing. In 2021, that was the metaverse, and SK Telecom was on its bleeding edge. The company launched its Ifland metaverse service in South Korea in 2021, amid the widespread belief that the pandemic would push users into virtual worlds. That was about the same time that Facebook renamed itself Meta (as in metaverse).

Of course, metaverse didn't pan out. SK late last year said it would shutter the Ifland platform, after reporting 4 million monthly active users (MAUs) on it in February 2023.

The reason? "As a global AI company, we aim to concentrate our capabilities on AI," according to an SK representative.

Aster is one part of SK's latest strategy. The only question now is when Aster and similar AI offerings will fall into Garner's "trough of disillusionment." There's also no telling whether they'll make it onto Garner's "plateau of productivity" promised land.

Article updated January 22 to correct Clark-Tan's title.

About the Author

Mike Dano

Editorial Director, 5G & Mobile Strategies, Light Reading

Mike Dano is Light Reading's Editorial Director, 5G & Mobile Strategies. Mike can be reached at [email protected], @mikeddano or on LinkedIn.

Based in Denver, Mike has covered the wireless industry as a journalist for almost two decades, first at RCR Wireless News and then at FierceWireless and recalls once writing a story about the transition from black and white to color screens on cell phones.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like