Rather than relying on pre-built tools, companies are placing AI directly into a learning loop, transforming messaging-based commerce from rigid scripts into flexible, personalized experiences.

Vietnam was among the early markets to shape social commerce. From manually closing sales via messages on platforms like Facebook and Instagram, the model has steadily evolved over nearly a decade and spread to countries such as Thailand, the Philippines, Malaysia, and even parts of Europe and the US.

Now, as generative AI reshapes how technology interacts with people, the domestic market is entering a new phase - Social Commerce 2.0.

At the center of this shift is a change in how businesses approach AI. Instead of simply deploying software, they are acting as “trainers”, guiding AI systems to advise customers, understand context, and ultimately close sales.

This evolution stems from a long-standing limitation of earlier automation systems. Traditional rule-based chatbots often created friction, forcing users to navigate rigid, pre-programmed flows just to access basic information.

According to Khoi Le, country director of Meta in Vietnam, the core issue lies in their structure. “With conventional chatbots, users must follow step-by-step instructions, like pressing numbers on a call center menu, which leads to a poor experience,” he said.

The emergence of AI agents powered by natural language processing has changed that dynamic. Instead of forcing users into fixed scripts, these systems can understand open-ended input and respond more fluidly.

Yet from a technical perspective, no AI model is complete at launch. Its effectiveness depends heavily on data flow and continuous human refinement.

This is where the “human-in-the-loop” approach comes into play. When AI encounters complex or unfamiliar requests, human staff step in to handle the conversation. The interaction is then captured and fed back into the system as high-quality training data.

Businesses review these conversations, identify gaps or inaccuracies, and fine-tune the model accordingly. Over time, the AI evolves beyond automated responses into a tool that actively drives revenue.

With sufficient data and proper calibration, these systems can analyze user behavior and make contextual recommendations. For example, a customer searching for coffee beans might be prompted with brewing equipment, increasing the overall order value.

Real-world data supports this shift. Sportswear brand Hadaki reported handling around 2,000 messages per day after integrating AI into its messaging system, doubling its previous capacity.

More notably, the company saw a 5% increase in order value thanks to personalized suggestions generated by AI. In another trial, an AI assistant was able to consult and complete a transaction within just 10 minutes of activation - without human intervention.

Such outcomes help explain broader trends. According to data from Kantar and Deloitte, AI-driven personalization tools can boost business profits by up to 20%. Meanwhile, 8 out of 10 Vietnamese users are open to interacting with businesses via messaging.

With 93% of companies in Vietnam already adopting AI tools on digital platforms - among the highest rates globally - the country is emerging as a leading market for conversational commerce and AI-driven discovery.

This momentum is one reason Meta has selected Vietnam as one of the first countries to roll out its latest AI transformation solutions.

Among them is the Business AI assistant on Messenger, launched on April 7, designed to operate 24/7 and streamline customer interactions while meeting growing user expectations.

 
Du Lam