South Korea's Biggest Carrier Gave Its AI Agents Employee Numbers, Job Titles, and a Retirement Plan
On October 26, SK Telecom published what it called its first AI Governance Report, and buried in the corporate-responsibility language is a genuinely unusual bet on how to manage autonomous software: treat it like staff. South Korea's largest wireless carrier, part of the SK Group conglomerate, has announced plans for a governance framework that would give its internal AI agents an employee number, an organizational affiliation, a job function and an authority level, managing their access to data and internal systems accordingly — a plan the company first announced in June and says it is now building into a formal, externally audited governance structure.
An HR file for software
The report sets out a single management standard that applies to SK Telecom's internal, employee-facing AI agents across four stages: development, testing, operation and termination — the same lifecycle a company might use to describe a product rather than a hire, except that the agents themselves are assigned the identity markers of one. Korean press reporting on the report confirms that four-stage label but doesn't detail exactly how sign-off, monitoring or retirement decisions work in practice — SK Telecom has not published operational specifics beyond the lifecycle framework itself.
SK Telecom paired the announcement with a concrete external validation: in October 2026 it secured ISO/IEC 42001 certification, the international standard for AI management systems, for its internal agent-management framework specifically, extending a certification it first won in 2024 for its customer-facing AI services. Chief Privacy Officer Cha Ho-beom, who oversees both personal-data protection and AI governance under one office, said plainly that "the importance of AI governance is growing even greater in the age of AI agents." The report traces back to values the company set in 2021 and governance principles it published in March 2024 under the name "T.H.E. AI," reinforced by an internal AI-governance portal that opened in September 2025 and an employee-education push called "Good AI" that launched in January.
A different philosophy than the one being sold this week
SK Telecom's move lands in sharp contrast to how Western enterprise vendors are framing the same problem. One day later, on October 27, IBM used its TechXchange conference to launch its own answer: an "Agent Identity" capability that works alongside HashiCorp Vault, establishing an authoritative identity record for each agent that Vault's existing credential controls then use to enforce access and revocation, plus a watsonx Orchestrate feature set meant to discover agents built on rival platforms including Amazon Bedrock, Microsoft Foundry and Google Gemini, and to contain any agent via sandboxing and a kill switch. IBM framed the stakes with numbers from Gartner: the average Fortune 500 company is projected to be running more than 150,000 AI agents by 2028, and today only 13% of organizations believe they have adequate governance over the agents they already have. IBM's pitch, like the security-credential approach Microsoft and other vendors have been selling agent-identity systems around in recent weeks, treats an agent fundamentally as a security object — something to be authenticated, scoped and revoked.
SK Telecom's framework starts from a different premise: an agent is closer to a worker than a workload. Giving it an employee number, a manager, a job description and a retirement date builds in something a pure identity-and-credential system does not — a native answer to who is accountable when the agent's output is wrong, because the org chart already says who it reports to. That approach may be harder to export outside a hierarchical Korean conglomerate culture built for exactly this kind of top-down structure, and SK Telecom's own disclosures leave plenty unsaid — no agent headcount, no detail on how performance is scored, no description of what actually triggers a termination. But as Gartner's own forecast suggests the agent population at large companies is about to explode by two orders of magnitude, SK Telecom is one of the first major enterprises to publish, audit and certify an answer to the organizational question rather than only the technical one — a bet that managing the people problem agentic AI creates might matter as much as managing the security problem.