SK Telecom prioritizes outcomes and cost-effectiveness in workplace AI

SK Telecom launched a companywide initiative in March to encourage employees to develop AI tools for their jobs[2]. The company says its agents help with preparing reports, analyzing mobile-network data and automating routine tasks[2]. AX Innovation 2.0 takes the effort further: SK Telecom says it will redesign organizational processes and collaboration around AI, rather than focusing only on employees’ individual use[1]. The shift puts the emphasis on how work is organized, as well as on the tools employees create.

More than 7,500 employees use AI agents each month, more than 4,000 have built their own agents, and about 1,500 agents are in regular use[2]. SK Telecom says it is prioritizing work outcomes and cost-effectiveness over simply expanding AI usage[3]. It evaluates agents individually for effectiveness and whether paying for their tokens makes sense[3]. The company says the goal of these assessments is to produce results, not simply to reduce the number of agents[3].

SK Telecom operates three workplace AI transformation programs: the AX Enabler Task Force, AX Sandbox and AX Catalyst[3]. The task force addresses companywide issues including data-access permissions, system integration and security[3]. In the first half of the year, the company tested for about three months a model in which one service-development employee worked with an AI agent on planning, development and verification[3]. It selected 10 people for its first cohort in September and plans to expand the program starting next year[3].

SK Telecom developed Z.E.Z., or Zero Exposure Zone, and filed a patent application for the technology[3]. It is designed to let AI process work without receiving sensitive original data, providing only necessary processing results; the company says it could support tasks such as medical-expense reimbursement[3]. A tool automatically adds descriptions and business context to data fields, helping users find information[3]. SK Telecom says answer accuracy on questions about refined data improved from around 40% to at least 80%[3]. However, it has not disclosed actual savings or companywide productivity improvements[2].