As generative AI rapidly penetrates the enterprise sector, financial institutions are expanding beyond customer-facing "smart customer service" to actively leverage AI in empowering internal business workflows. At the "2026 Taiwan Generative AI Applications Hackathon" - guided by the Administration for Digital Industries (ADI), Ministry of Digital Affairs (MODA), organized by DIGITIMES, and supported by Amazon Web Services (AWS) with international AI technology expertise, TS Holdings presented "Financial Innovation" as its challenge theme. Addressing the high-frequency pain points of data consolidation and analysis within banking operations, the company designed the "Smart Data Presentation Tool" challenge to explore AI application models that deliver tangible operational efficiency gains through cross-disciplinary experimentation.
Empowering Business Operations: Shift from B2C to B2E
Steve Sun, Chief Information Officer at TS Holdings, noted that business units across the bank frequently need to prepare high-level reports covering industry trends, peer dynamics, competitive analysis, and customer profiles. These tasks involve retrieving, consolidating, cross-checking, and interpreting large volumes of internal and external data before distilling the findings into concise, actionable insights for management decision-making. In the past, such work relied heavily on manual processes and tools like Excel, making it both time-consuming and labor-intensive. Therefore, the challenge specifically guided participants to integrate AI technologies to automate data retrieval, organization, analysis, and recommendation generation. Moving beyond the company's historical focus on consumer-facing (B2C) AI applications, the hackathon served as a vital testing ground for TS Holdings to validate internal employee-facing (B2E) AI empowerment tools.
Interdisciplinary Innovation: Automated Verification and Executive-Ready Reporting
In the "Financial Innovation" challenge category, a team comprising students from the Department of Economics at Feng Chia University and the Department of Psychology at Fu Jen Catholic University earned top honors. Sun analyzed that the team's winning advantage lay in its comprehensive "automated data verification mechanism" established post-retrieval, which effectively prevented erroneous data and AI hallucinations from propagating down the analytical pipeline. The team flexibly integrated AI tools - including the Claude model accessed via Amazon Bedrock, AWS's fully managed service offering a choice of high-performing foundation models from leading AI companies - for data cross-validation and content generation. Furthermore, through prompt engineering, they refined the output structure into concise, well-formatted reports tailored to executive reading preferences. TS Holdings highly praised the solution's end-to-end execution, noting that its workflow - from data preprocessing and logical analysis to final presentation - was meticulously designed around real-world financial scenarios, showcasing the innovative caliber of cross-disciplinary talent.
Amazon Bedrock provides managed foundation model services, allowing development teams to access and utilize diverse models within a single unified environment. This reduces the overhead of constructing and maintaining dedicated model infrastructure while accelerating rapid prototyping and tool integration. For complex enterprise applications requiring concurrent execution of search, reasoning, and content generation, such cloud model services offer high architectural configuration flexibility.
Modular AI Agents and Human-in-the-Loop Governance
Looking ahead to enterprise deployment, TS Holdings evaluates initiating implementation in targeted vertical scenarios, such as competitive analysis for the finance department. This allows the team to sequentially validate data sources, verification mechanisms, model inference performance, and presentation quality before iteratively optimizing based on user feedback. Over the longer term, this architecture can be further deconstructed into independent "AI Agents" handling search, data cleansing, integration, and chart generation. As technology matures, these modular agents can be flexibly extended to other core functions, such as credit reviews, significantly increasing tool reusability. Given the financial sector's stringent standards for data precision and regulatory compliance, actual deployment will integrate a "Human-in-the-loop" review framework, with business domain experts conducting final validations to balance AI efficiency with professional rigor.
Sun emphasized that the competition highlighted the immense possibilities of bridging cross-disciplinary talent with financial practice. He observed that as the barrier to adopting AI tools lowers, "the ability to deeply understand financial workflows, grasp operational pain points, and define the correct problems" has become the paramount factor in determining AI success. This necessitates close synergy between the domain knowledge of frontline business teams and the technical prowess of IT departments. Moving forward, TS Holdings aims to maintain deep engagement with participating teams, utilizing internship programs to give emerging talent hands-on exposure to authentic financial operations. Through hackathons and collaborative platforms, the enterprise gains fresh problem-solving perspectives while participants gain direct insights into industry demands, building a rich foundation for Taiwan's fintech talent pipeline.
