As generative artificial intelligence (AI) matures rapidly, the challenges for an organization to adopt AI are not only limited to tool selection or technological effectiveness, but how to avoid fragmented applications, minimize governance risks, and integrate AI into daily workflows in a way that enables sustainable organizational capability. For postal services with diverse business types and a national service network, the lack of a unified architecture and governance mechanism will not only make it difficult to scale across the organization, it may also create information disparities, increase administrative burden, and even affect the stability of overall organizational operations.

Since the initial implementation of generative AI, Chunghwa Post has adopted a "platform-based, governance-first" strategy, establishing an internal generative AI platform as the common infrastructure for various AI applications. A centralized architecture and standardized service design enables the gradual expansion of AI applications in a controllable and traceable manner, preventing fragmented tool adoption while laying a stable and consistent foundation for future application development. Moreover, to transform technological capabilities into concrete and easy-to-use services, Chunghwa Post has integrated services from a user experience perspective and created the intelligent virtual character “Ban-won” as a common symbol and operating interface for internal AI services. This allows our employees to gradually build trust and develop consistent usage habits through a unified user experience, making AI a reliable tool for supporting their daily work. The platform and related application training were launched in 2025. In addition, the "Postal AI Hackathon" is also organized to nurture AI talent and encourage employees to unleash their potential. Please see 4.1.2 Building Growth Momentum for details[y.

To ensure that the responses are based on internal company data and to maintain accuracy and traceability, the postal service’s internal generative AI platform combines Large Language Models (LLM) with its existing knowledge management system and incorporates Retrieval Augmented Generation (RAG) technology. The system can provide decision support and operational guidance within a clearly defined data scope, serving as a reliable and trusted business support system. The platform’s first application, the AI service of “Teller Q&A” service, focuses on supporting frontline service operations, enabling tellers to instantly access operational guidance and reference information through conversational interactions in real time. This not only improves information retrieval efficiency, but also facilitates the systematic updating of a large number of old and scattered documents, making knowledge a usable and trustworthy digital asset again, which further shortens education and training time, and improves service consistency and risk control capabilities.

The interface is designed with a predefined FAQ menu that addresses common customer service inquiries. Each response includes links to its information sources to enhance information transparency and user trust. At the same time, the system also incorporates a feedback mechanism to collect employees’ actual usage experience, which serves as an important basis for subsequent optimization of model accuracy and knowledge content, gradually forming a continuous improvement cycle over time.

In addition to frontline services, the platform also launched "Administrative GPT" as an administrative assistant, supporting text generation, content summaries, multilingual translation, document processing, and speech-to-text to assist employees in handling routine administrative tasks, allowing them to be relieved from repetitive work and allocate more effort to higher-value activities, such as analysis, communication, and coordination to further enhance the overall effectiveness and efficiency of organizational operations. With the establishment of an internal generative AI platform, the postal service has moved from a single application to a governance-based AI service architecture. This approach not only allows room for future expansion and more advanced application deployment, but also strengthens organizational digital resilience, creating a sustainable intelligent transformation model that aligns efficiency, governance, and sustainable development.