วิเคราะห์-วิจารณ์: KBTG กับความจริงของ "Agentic AI" ที่ยังไม่ได้ถูกพูดในสัมภาษณ์

The analysis of KBTG and the reality of Agentic AI, which has not been discussed in interviews, highlights several key points. KBTG's Executive Chairman, Krithir Rojnapong, has taken over as the primary leader following his appointment as advisor to the board of KBank. This interview marks the first official strategy communication from the new leadership, emphasizing stakeholder management beyond mere sharing of experiences.

The discussion touches on concrete data points, such as coding speed increasing by 200-300%, but the overall software development life cycle only sees a 10-15% reduction in time. This is because coding is just one component of the requirement, design, coding, testing, and deployment process. Techsauce reports that AI contributed to 21 million lines of code (around 10% of total code) in 2025, with a target of 15% by 2026.

This example showcases that productivity gains are not exaggerated claims but rather realistic improvements. KBTG acknowledges the high token costs associated with AI coding and warns that organizations must optimize their choice of models for specific tasks to avoid negative consequences. Human-in-the-loop and observability are emphasized as crucial in managing AI agents, aligning with global AI governance practices focusing on data access control, agent identity, and kill-switch mechanisms.

The interview also highlights that success in AI adoption is not solely determined by model quality but by the approach and strategy employed. KBTG's approach of building AI solutions in-house, despite facing a low success rate (only ~33%) compared to partnerships with vendors (67%), is a strategic choice backed by their strong resources and track record.

The interview offers valuable advice on focusing on pain points and starting with realistic expectations, rather than solely chasing high ROI targets. However, it is essential to consider the potential risks and challenges associated with building AI solutions in-house, such as higher failure rates and the learning curve for younger generations entering the workforce.