Language was never the problem: Inside SEA’s real AI adoption gap

Kai Yong Kang, a Partner at GenAI Fund, argues that language barriers are no longer the main obstacle to AI adoption in Southeast Asia. Instead, enterprises face trust issues when it comes to deploying AI systems securely and at scale within their business processes. The real gap lies in whether an enterprise can trust an AI system to execute tasks accurately, securely, and efficiently.

This distinction is central to the conversation surrounding AI adoption in the region. Most companies fail to achieve true localisation by stopping at language translation. Successful localisation requires addressing language, culture, and operations. Kai's firm conducted an Agentic AI Build Week, which attracted over 3,000 AI builders and resulted in 400 solutions for enterprises like KFC Vietnam and Tasco.

Through this initiative, they gained insights into the challenges faced by enterprises in Southeast Asia. A notable example is the Twohearts team, which developed an agentic ordering system for KFC Vietnam that pulls live menu data, applies vouchers, confirms orders, and hands off complicated tasks to human staff. This illustrates the importance of direct experience with workflows in achieving effective localisation.

Kane also emphasizes the need for shared responsibility in addressing the language-data gap in Southeast Asia. While governments, universities, technology companies, and enterprises all have a role to play, the primary issue goes beyond data availability. It lies in establishing repeatable mechanisms that connect datasets to real-world problems.