Artificial intelligence (AI) can open unprecedented opportunities for Vietnamese businesses, improve operational efficiency, reduce costs, increase revenue, and increase competitiveness on the global stage.
Vietnam has identified AI as a key strategic technology to raise its global position. Under Poliburo's resolution 57-NQ/TW on breakthroughs in science, technology, innovation and national digital transformation, Vietnam aims to become the top three countries of ASEAN countries and the world's top 50 in AI development by 2030. The government has listed AI as a top priority in its national strategy on emerging technologies.
Recently, Vietnam has adopted the first legal framework of AI, adopting laws relating to the science, technology, innovation and digital technology industry.
Phan Thi Thanh Ngoc, director of AI consulting for VNPT AI, a subsidiary of VNPT Group, not only has adopted AI quickly, but he also has built his own ecosystem and platform to meet domestic needs and compete with global tech giants.
VNPT began research and development on AI more than five years ago and currently offers a wide range of AI products for multiple customer segments.
In the public sector, VNPT uses AI to enhance digital governance. For individuals, the company develops virtual AI assistants. In the enterprise sector, AI-powered solutions help businesses increase revenue, improve profits and improve customer experience.
One outstanding product is VNPT's AI-based EKYC system. It ranks in the top 10 in the world. VNPT has developed a large-scale language model (LLM) optimized in Vietnam, which is currently leading the national rankings. All of these technologies are fully developed in-house and demonstrate Vietnam's ability to compete internationally in AI.
Tech Giant VNG makes a comprehensive investment in AI across its infrastructure, platforms and applications. In June 2024, Green Node, a cloud subsidiary, launched Southeast Asia's first large-scale AI cloud. Within six months, we made significant profits due to high market demand.
Vietnam is now one of the first three Southeast Asian countries to develop its own LLM, with Zalo's Kilm model achieving comparable performance to the GPT-4 in just six months. In particular, about 20% of Zalo's 75 million users (15 million people) use AI-powered features such as translations and AI avatars every day.
AI is also transforming banking operations. Previously, loan processing required large teams to manually enter and cross-check the data. With AI document extraction from GreenNode, banks are currently processing millions of documents each year, saving billions of Vietnamese dong in operational costs. Processing times have been improved by at least 30%, allowing banks to efficiently serve more customers.
Vu Thanh Tung, product director at GreenNode, observed that both the government and corporate sectors are accelerating practical AI applications.
In government, AI automates resume processing. Insurance helps to digitize and streamline claim handling and other document-rich processes.
Despite these advances, the widespread adoption of AI in Vietnamese companies faces several barriers, including infrastructure costs, deployment challenges and limited awareness.
“If business leaders lack a broad understanding and long-term vision to value AI, they may be hesitant to invest in it, and some may be worried that AI will replace their work,” the NGOC said.
She emphasized the need for a clear investment strategy and focusing on safety, security and privacy. “Vietnamese companies should prioritize “Vietnamese-made” AI solutions tailored to local needs,” she advised.
From a technical standpoint, Tung pointed to two important data and security concerns.
“Biometric authentication has led to deep-fark attacks as a new threat, while effective AI deployments rely on having clean, well-organized, centralized data,” he said.
To address this, GreenNode is developing Data Bloom, a platform that integrates fragmented data sources. This allows non-technical business teams to access and analyze data in their AI applications.
“These AI models don't have to be ChatGpt. There are open source, low-cost alternatives like Deepseek, Qwen, and Llama,” added Tung.
