KT said on the 2nd that it will move to target the corporate artificial intelligence (AI) market by unveiling Agentic Fabric, an AX (AI transformation) operating system optimized for corporate environments, at MWC 26, the world's largest mobile communications expo that opened in Barcelona, Spain.
Agentic Fabric is an enterprise AI operating system (OS) designed to directly perform corporations' core tasks, unlike existing Generative AI, which has been confined to auxiliary functions such as document summarization and search. Its features include autonomy that minimizes human intervention, alignment with corporations' goals, values, and policies, and execution capability that goes beyond simple answers to actual task performance.
KT said Agentic Fabric weaves together diverse AI technologies and agents like fabric to carry out the full range of corporate work. It resolves various AX hurdles that corporations face when applying AI to core operations, including complex system integration, data security, memory discontinuity, and unpredictable decisions.
Specifically, it is designed with a total of five layers: an experience layer that creates and manages agents with a single user experience (UX) from development to operations; an intelligence layer that handles non-deterministic reasoning and rule-based task execution; a context layer that accumulates and assets corporations' domain knowledge and work experience; an execution layer that connects internal and external systems and various tools to lead to actual execution; and a governance layer that controls agents based on security, policy, expense, and audit.
It also adopts an architecture that separates the management and control domain from the execution domain, so core data is managed securely while AI models and technologies can be flexibly expanded as needed. It additionally supports various IT environments and regulatory conditions, including software as a service (SaaS), on-premises deployment, and hybrid.
In the AX platform zone within the KT booth, the exhibition centers on cases of Agentic Fabric operating in real corporate environments. Visitors can directly observe the entire process in which AI's judgment leads to actual task execution through Agentic Fabric.
A representative example, "agent orchestration," showcases a multi-agent operating method in which an orchestrator manages the entire process under a single task goal and multiple specialized agents, each focused on roles such as diagnosis, analysis, and optimization, collaborate. It is characterized by analyzing the user's query, establishing an execution plan, breaking it down into subordinate tasks to assign to each agent, and completing the final deliverable after a process of synthesis and verification of the results.
The strength of this architecture is that it can break down and complete complex corporate tasks through collaboration among multiple specialized agents rather than a single AI. For example, when a problem such as "degraded network quality in a specific area" is detected, the orchestrator distributes the work to relevant agents to resolve the issue and compiles the results into a single report. A KT official said, "The success experiences accumulated in this process are then used as the basis for agents' judgments, enabling faster and more accurate responses when similar situations arise in the future."
It also introduces "K RAI assessment" for responsible AI operations. Visitors can see a dedicated RAI assessment agent that automates the entire process by independently developing an assessment plan, conducting red-team testing, performing in-depth analysis, and generating a report. Based on an AI risk classification system, it precisely evaluates 11 risk areas and even proposes comprehensive risk-mitigation solutions, providing an environment in which corporations can confidently adopt and operate AI.
Oh Seung-pil, head of KT's Technology Innovation Division (vice president), said, "Agentic Fabric is a corporate AI OS designed to control and execute AI on a single operating system even when each corporation's domain and work environment differ. As AI transitions into a principal actor in corporations' core tasks, it will improve the structural limitations of corporate AX," adding, "Based on this, KT will lead the market as an agentic AI partner that substantively completes corporate AX."