SK Telecom's AI Native Strategy at MWC26: A CEO-Level Briefing for Executives
SK Telecom's CEO Jung Jai-hun set a clear agenda in Barcelona: compress customer value and AI into one operating model and rebuild the company around it. The goal is ambitious-help position Korea among the top three AI leaders-yet the levers are concrete: overhaul systems, convert operations to autonomous modes, and scale infrastructure to 1GW-class AI data centers.
As Jung put it, SKT is at a "golden time of transformation," where customer value innovation meets AI innovation. The message to leaders is simple: re-architect the core, not the edges.
What's Changing-Beyond the Slogans
- End-to-end IT rebuild for AI: Sales, line management, and billing move to AI-optimized, integrated stacks to enable dynamic plan design and instant personalization.
- Security by default: Company-wide Zero Trust with strict authentication, access control, network segmentation, and AI-driven monitoring. See the NIST Zero Trust guidance.
- Autonomous network operations: AI-directed wireless quality management, traffic control, and facility operations-aimed at consistent quality and ultra-low latency with AI-RAN.
- Simpler commercial models: Pricing, roaming, and membership restructured for clarity; personalized packages offered automatically.
- Unified customer experience: A single integrated AI agent connects T world, T Direct Shop, and other touchpoints, reading daily patterns to deliver timely recommendations.
- AI-augmented service: AICC equips every representative with guidance for faster, more accurate support; offline stores shift from selling to advisory and ongoing curation.
- Deeper segmentation: "AI Personas" turn behavioral data into conversational profiles so teams can understand needs and respond in plain language.
- Device-level agent: A. phone (A-DoT phone) evolves into a true agent-auto-summarizing calls, updating schedules, connecting to services, and taking actions.
Infrastructure: From GPU Cluster to 1GW-Class AIDC Network
SKT will build hyperscale AI data centers across Korea, targeting aggregate capacity above 1GW and establishing an Asia-leading AIDC hub. In partnership with OpenAI, a new AIDC in Korea's southwest will extend a nationwide AI backbone.
With SK hynix, SK Ecoplant, and SK Innovation, SKT is integrating the full stack-construction, cooling, servers, energy, and operations-to drive cost efficiency. For leaders watching sustainability and grid constraints, reference points like IEA analysis are useful context.
- Petasus AI Cloud: Virtualization applied to Haein (GPUaaS) and offered globally this year.
- Sovereign foundation model: Scaling from 519B to over 1T parameters, adding multimodal capabilities (images, voice, video) starting in H2.
- Manufacturing AI package with SK hynix: Real-time process analytics to cut defects and lift equipment efficiency-delivered as infrastructure, model, and solution.
Why This Matters for Executives
- Customer economics: Personalization at design-time and run-time improves perceived value, ARPU uplift, and retention without bloating the catalog.
- Operating leverage: Autonomous network operations shift effort from manual oversight to exception handling, improving quality and margins.
- Strategic control: Sovereign models and domestic AIDC footprint reduce external dependency, protect data, and stabilize cost trajectories.
- Ecosystem pull: Collaboration with chip, construction, and energy partners concentrates capability and accelerates time-to-impact.
Execution Guardrails (So You Don't Pay the AI Tax)
- Data governance first: Consent, lineage, and retention policies must be embedded in every agent and system call-especially across unified CX and AICC.
- Zero Trust everywhere: Treat every request as untrusted; verify continuously. Automate audit trails.
- Resilience by design: Plan for model drift, degradation, and failover. Keep human-in-the-loop for high-risk actions.
- Energy and siting strategy: Lock in cooling and electricity plans early; track PUE and grid constraints per site.
- Integration discipline: Sequence modernization-don't wire AI agents onto brittle legacy processes without refactoring the core.
Metrics That Signal Real Progress
- Customer: NPS, churn, ARPU uplift from personalized plans; resolution time and CSAT in AICC.
- Network: % autonomous incidents handled, latency variance, quality-of-experience indices by segment and location.
- Infrastructure: Provisioning lead time, GPU utilization, PUE, cost per trained/inferenced token.
- Model: Task accuracy, hallucination rate by domain, safety scorecards, time-to-fix for regressions.
- Manufacturing AI: Defect reduction, throughput gain, OEE improvement, payback period by line.
90-Day Action Plan for Operators and Enterprise Leaders
- Map the customer journey to one agent: Identify 3-5 high-friction moments and connect them under a single orchestration layer.
- Stand up a Zero Trust baseline: Migrate one critical workflow end-to-end with continuous verification and policy enforcement.
- Build the AI operating model: Define model ownership, evaluation gates, incident response, and release cadence.
- Pick one industrial use case: Deploy a manufacturing AI pilot with clear data contracts and ROI targets.
- Lock infra decisions: Choose a reference stack (virtualization, observability, MLOps) and a site plan aligned to energy and cooling realities.
Inside SKT's Work Culture Shift
Jung highlighted a ground-up reset of how teams work. An AX (AI Transformation) Dashboard tracks adoption by department and individual, while an AI Board backs teams with funding and support.
Employees use an AI playground to create agents without code; more than 2,000 agents are live across marketing, legal, and PR. Company-wide upskilling is in motion to make AI a daily habit, not a side project.
What to Watch Next
- Integrated agent performance: Conversion and retention impact once T world and T Direct Shop experiences unify.
- Autonomous ops at scale: How quickly incident response and latency stabilize across dense urban zones.
- Model upgrade to 1T and multimodality: Gains in Korean-language nuance, voice and video comprehension, and enterprise safety.
- Manufacturing AI outcomes: Measurable yield and OEE gains across early deployments with SK hynix and partners.
Further Reading for Leaders
CEO Jung summed it up well: AIDC as the heart, hyperscale models as the brain. Combine them with customer-first design and disciplined execution-and you get a company built for the next decade, not just the next quarter.
Your membership also unlocks:
