SK Group Pilots Agentic AI for Autonomous Manufacturing, Targeting 2027 Commercialization

SK Group has started a pilot project for 'autonomous manufacturing,' using AI to detect and resolve equipment failures on its own.

This initiative is driven by three key factors: the semiconductor supercycle, the availability of suitable open technology, and alignment with SK's group-wide AI strategy.

The goal is to eliminate unplanned downtime, which could boost effective production capacity by over 10% without new facility investments.

SK Group has begun testing a groundbreaking AI system designed to create a 'self-driving' factory.

This pilot project, announced on March 12, 2026, combines SK Telecom's small large language model (sLLM) and SK AX's agentic AI software with Bosch Rexroth's industrial equipment. The system aims to autonomously detect, interpret, and act on rail system anomalies, with the ambitious goal of achieving nearly 100% uptime by virtually eliminating unplanned downtime. The technology is slated for commercialization in 2027 after a proof-of-concept phase.

So, why is this push for 'autonomous manufacturing' happening now? Three key trends are converging.

First is the semiconductor supercycle. With soaring demand for AI chips like HBM, any production bottleneck becomes extremely costly. South Korea's exports in February surged, led by a 134% jump in semiconductors. In this environment, reducing downtime is equivalent to a direct increase in production capacity, making the return on investment for automation incredibly high.

Second, the technology is finally ready. Bosch Rexroth's ctrlX OS is built on open standards like Linux, making it structurally easy to integrate external AI software. This allowed SK to 'layer' its AI agent on top of the existing system, transforming it from a simple monitoring tool into an intelligent decision-maker that can cross-reference manuals and past cases to decide on actions.

Third, it perfectly aligns with SK Group's overarching AI strategy. This includes SKT's 'Sovereign AI' initiative and the 'Manufacturing AI Cloud' partnership between SK Group and NVIDIA, which aims to build digital twins of factories. This pilot is a practical application of that grand vision, bringing high-level AI concepts down to the factory floor.

The quantitative impact is significant. Industry data suggests that unplanned downtime often accounts for over 10% of lost production time. Eliminating this 10% loss translates to an 11.1% increase in effective operating time—a massive efficiency gain without building new facilities. This project marks a potential turning point for manufacturing, shifting from passive monitoring to proactive, autonomous operation.

Agentic AI: An AI system that can proactively take actions and make decisions to achieve goals, rather than just processing data or responding to prompts.
LMS (Linear Motion System): Equipment that moves parts or tools along a straight line, common in automated assembly and production lines.
sLLM (small Large Language Model): A more compact and efficient version of a large language model, designed to run on smaller devices (like factory equipment) rather than in large data centers.