SK Telecom shifts its AI push from employee tools to company-wide workflows

SK Telecom is moving its internal AI effort beyond tools that help employees work alone. On November 1, 2026, it announced AX Innovation 2.0, a program to redesign workflows and collaboration around AI. The effort combines organizational changes with technologies intended to make company data safer and more useful to AI agents.

The company’s stated shift is from individual convenience to practical gains across the organization. It identifies data access, connections between systems and security as obstacles to that transition. SKT describes an “AI transformation J-curve”: efficiency can initially fall as employees retrain, workflows change and data is reorganized, before productivity improves. That is the rationale for the initiative, not a measured result of it.

More than 4,000Employees who have built AI agents

SK Telecom says more than 4,000 employees have tried building AI agents.

1,500Agents used regularly

The company reports 1,500 AI agents in regular internal use.

Three programs connect experiments to company support

SKT’s implementation has three parts, covering shared obstacles, workflow experiments and support for employees already applying AI to their jobs:

• AX Enabler TF is a company-wide task force responsible for recurring difficulties in AI use. Its remit includes training, support, data and security infrastructure. It also covers management changes, including introducing employee-style identification numbers for AI agents and establishing an evaluation framework.
• AX Sandbox starts with redesigning work around AI rather than preserving existing procedures. In the first half of the year, a service-development team experimented with one employee handling planning, development and validation alongside agents. In the second half, SKT is extending experiments across organizations to explore collaboration among multiple employees and agents.
• AX Catalyst supports employees who are already actively using AI in their work. Ten people were selected for the first cohort in September. SKT supports their AI-model usage costs and technical and data connections, while asking them to share agents and experience with colleagues and identify obstacles to wider adoption.

The programs are designed to connect. When Catalyst participants encounter problems involving data, permissions, systems or security, they work with the Enabler task force to resolve them. SKT’s aim is to prevent other employees from repeatedly running into the same barriers, rather than leaving each team to solve them independently.

Useful data without exposing the originals

On the technology side, SKT says it developed Zero Exposure Zone, or Z.E.Z, and filed a patent application. The system keeps sensitive original data, including personal information, in a separate area. AI receives only the processing results needed for a task, such as metadata—information describing data—comparison tokens or calculated values, rather than the underlying records.

A separate tool, Data Desk, addresses meaning rather than exposure. SKT says company data has traditionally been organized around systems and developers, which can make its business purpose difficult for AI to interpret. Data Desk automatically generates descriptions and work context so agents can find and use the information they need.

Mapping the work before assigning an agent

SKT is also developing Work Map to make individual and organizational workflows understandable to AI. It starts with existing work materials and goals, uses AI to infer how work flows, then builds a map through conversations with employees. The intended outputs are standard procedures for people and AI working together, plus an inventory of tasks where agents could be applied.