
Everyone’s AI brings Korea’s telecom and messaging platforms into consumer agents
The AMW Read
The planned rollout meaningfully expands consumer-agent distribution, while the domestic-model usage mandate links nationwide adoption to sovereign-AI policy; execution and adoption remain unproven.
Everyone’s AI brings Korea’s telecom and messaging platforms into consumer agents
SK Telecom, Kakao and KT plan October beta releases for Everyone’s AI, a South Korean government-backed initiative, followed by formal launches in December 2026. Three consortia will provide general-purpose chatbots, public-service agents and specialized services developed with startups. Basic functions will be free, while advanced services may carry separate business models. Each consortium targets 5 million monthly active users by year-end; these are adoption goals, not reported results.
The market significance lies in distribution and domestic-model demand. Kakao plans to make reservations, applications and payments accessible through KakaoTalk. SK Telecom is developing agents that plan, execute and check tasks, including calling businesses or government offices without APIs. KT’s Ieum will connect search, comparison and public-service tasks with partner services and telecom channels. Requiring domestic models to account for at least 50% of AI usage turns consumer distribution into a sovereign-AI demand mechanism. The government also plans 250 billion won in funding for 2027 and infrastructure support equivalent to roughly 2,000 Nvidia B200 GPUs; neither is described as already delivered.
For builders and investors, the concrete opportunity is integration with established messaging, telephone and service channels. Kakao’s proposed agent marketplace could give outside developers a distribution route, but the key diligence question is whether agents can complete authorized transactions reliably. The article identifies an unresolved constraint: using academic and income records to submit public-benefit applications requires rules defining private operators’ access to administrative data. Launch schedules and user targets therefore do not yet establish dependable execution or sustainable paid demand.



