SK Telecom is defining “AI Native” as a rebuild rather than a collection of features. At MWC 2026, the company described plans spanning sales, billing, networks, customer service, retail, data centres, a sovereign foundation model, manufacturing solutions, and internal culture.
The scale is ambitious: more than one gigawatt of AI data-centre capacity across Korea, a model expanding from 519 billion to over one trillion parameters, and an operating environment in which more than 2,000 internal agents are already used.
The strategy’s coherence is more important than any single number. SK Telecom is attempting to connect infrastructure, models, applications, and governance as one corporate transformation.
Legacy Telecom Systems Become AI Infrastructure
Sales, line management, and billing systems will be redesigned to support personalised plans and memberships based on customer usage. These platforms are core operational records, so modernisation can change service design as well as technical efficiency.
The company plans Zero Trust controls including stronger authentication, access management, segmentation, and AI-based monitoring. That is necessary because more connected data and automation increase the consequence of compromised identity or agent access.
Autonomous network operations will apply AI to wireless quality, traffic control, and equipment. AI-RAN technology is intended to improve speed and latency.
Customer data use needs clear boundaries. Personalisation across billing, network behaviour, and retail can be useful, but customers should understand how recommendations are formed and be able to challenge errors.
The system also needs safe modes when automated network or billing decisions become uncertain.
One Agent Connects Customer Touchpoints
SK Telecom plans an integrated agent across its T world portal and T Direct Shop, with tailored suggestions based on daily usage. Contact-centre staff will receive AI assistance, while store employees use insights after customer visits.
This could reduce the familiar problem of customers repeating information across app, call centre, and shop. A shared context layer can preserve memory across channels.
The risk is that a single agent also concentrates errors and permissions. A wrong profile inference may follow the customer everywhere. Each channel needs access only to relevant information, with visible correction and escalation.
SK Telecom’s “AI Personas” will analyse digital behaviour across customer segments. Governance should test whether segmentation creates unfair offers, intrusive targeting, or assumptions that customers cannot see.
Useful measures include resolution quality, plan suitability, complaint rates, channel handoffs, employee verification time, and customer control over personalisation.
National Infrastructure Supports the Corporate Plan
SK Telecom aims to build hyperscale AI data centres exceeding one gigawatt and attract global investment. It already operates the Haein GPU cluster and Petasus AI Cloud and plans wider cloud distribution.
The sovereign foundation model will gain multimodal image, voice, and video capability. Parameter count is a capacity indicator, not proof of useful performance. Evaluation on Korean language, telecom workflows, safety, latency, and operating cost will matter more.
The company is also working with SK hynix on a manufacturing package that analyses process data to reduce defects and improve equipment efficiency, offered as infrastructure, model, and solution.
Internally, an AI Board, transformation dashboard, and no-code “playground” aim to make adoption measurable and accessible. More than 2,000 agents across marketing, legal, and public relations create an immediate need for inventory, permissions, and evaluation.
SK Telecom is tying its own transformation to Korea’s ambition to become a leading AI power. Execution will require enormous capital, energy, talent, and coordination.
The plan is compelling because it does not pretend that a model alone makes a company AI-native. Billing must change, networks must change, customer workflows must change, and governance must connect them.
If those elements remain aligned, SK Telecom could become a genuine example of AI as an operating model. If they fragment, the impressive infrastructure risks becoming a collection of costly parallel programmes.
