[E-Daily Reporter Yun Junghoon ]KTCorporation(030200)is significantly upgrading Woori Bank’s remote financial consultation system and accelerating its push into the “Agentic AICC” market, where AI performs tasks autonomously. The company plans to build a next-generation consultation environment where AI goes beyond simple guidance to handle actual financial transactions.

KTCorporation announced on the 31st that it has been selected as the contractor for the “AI Chatbot and Consultation Bot Reconstruction” project commissioned by Woori Bank and is set to begin full-scale system development.

The core objective of this project is to upgrade Woori Bank’s chatbot and consultation bot infrastructure and to organically integrate “AI Bankers” and AI agents, thereby expanding the scope of AI utilization to include not only inbound and outbound interactions but also in-branch consultations.

In particular, KTCorporation will apply its new solution, “Agent Connector,” to this project for the first time. Agent Connector is an orchestration (integrated management) technology that connects different channels—such as chatbots, consultation bots, and AI Bankers—into a single system, ensuring that the context of a previous conversation is maintained even when a customer switches channels.

Multi-LLM Routing… Optimizing Security and Token Costs

According to a briefing held that day, KTCorporation announced that for this Woori Bank project, it will not rely solely on its proprietary large language model (LLM), “Mideum,” but will instead adopt a “multi-LLM routing” strategy tailored to the client’s environment.

Seo Dong-cheol, a manager at the AX Tech Division of KTCorporation’s AX Business Division, explained, “Using a single model increases the burden of token costs,” adding, “We are configuring models of various sizes for each agent task and supporting the automatic selection of the model with the lowest token cost.”

In particular, considering the financial sector’s strict network segregation regulations and data security requirements, the system is built using an “on-premise” approach, where the model is deployed directly on Woori Bank’s internal servers. This approach curbs unpredictable increases in token costs and alleviates concerns about data leaks.

Speech Recognition and Synthesis Utilize KTCorporation’s Proprietary Technology… Phase 2 Implementation Targeted for November

Unlike the LLM domain, where orchestration with open-source and external models is flexibly applied, the speech recognition (STT) and speech synthesis (TTS) domains are fully powered by KTCorporation’s proprietary technology.

Manager Seo explained, “STT and TTS utilize KTCorporation’s proprietary technology; specifically, for TTS, we’ve enhanced real-time responsiveness by applying a technique that optimizes and trims sentence segments.”

Following the launch of the Phase 1 system last July, KTCorporation plans to complete the Phase 2 enhancement this coming November. The goal is to increase the coverage of customer service tasks from the current 30% to a maximum of 60%.

However, the company plans to take a phased approach for high-risk tasks, such as asset transfers and loans.

Heo Yang-seok, Team Leader of the AX Business Division at KTCorporation’s AX Business Headquarters, predicted, “Based on analyses of the customer environment and business operations, high-risk tasks will require a scenario-based approach for the time being,” adding, “A full transition to an agent-based system will likely be possible around 2028, once technical maturity and safety have been verified.”

Team Leader Heo added, “When promoting B2B AX (AI transformation) projects, rather than imposing a heavy, monolithic platform on the entire system, we will adopt a modular approach tailored to the client’s environment to ensure a smooth transition to an agent-based AICC.”

Kim Won-tae, Executive Vice President and Head of the Public/Financial Business Division at KTCorporation’s ENTERPRISE Division, said, “Based on our experience in building and operating AICC systems for four of the five major commercial banks in Korea, as well as our AI technology capabilities, we will lead the innovation of customer service at Woori Bank and the expansion of financial AX.”