Singaporeans embrace AI shopping agents, but demand safeguards and oversight

A new study on agentic AI—software agents that can make decisions and take actions on behalf of users—reveals a population caught between curiosity and caution. Global Payments released research on Wednesday surveying seven countries, including the United States, China, and the United Kingdom, to understand how willing consumers are to hand over purchasing decisions to artificial intelligence. The results paint a picture of Singapore as a market intrigued by the technology but deeply skeptical of its autonomy.

Fifty-nine percent of Singaporeans say they would be open to letting an AI agent shop for them. That's a substantial majority, suggesting real appetite for the convenience such tools could offer. Yet the numbers narrow sharply when the question becomes more specific. Only 23 percent have actually used an AI shopping agent. And when researchers asked how far people would go—allowing an agent not just to browse and recommend, but to shortlist options and process payment entirely without human intervention—just 15 percent said yes. The gap between openness and trust is the story the data tells.

The vision that payment companies and AI startups are building toward is one where people authorize agents to handle a wide range of transactions on their behalf. The applications are varied: switching insurance policies, booking travel, managing subscriptions, comparing prices across vendors. The technology promises to save time and reduce friction in commerce. But the survey suggests consumers see a meaningful difference between an agent that helps them decide and an agent that decides for them.

What stands out most sharply in the research is Singapore's demand for government oversight. Thirty-four percent of Singaporean respondents want a government-backed AI safety certification—the highest proportion among all seven countries surveyed. This figure suggests that Singaporeans are not simply nervous about the technology; they are looking for institutional reassurance. They want a third party, specifically the state, to vouch that an AI agent is safe to trust with their money and their choices. No other country in the study showed this level of appetite for regulatory certification.

The study exposes a fundamental tension in the rollout of agentic AI for consumer commerce. The technology's promise lies in its ability to act autonomously, to make decisions faster and more comprehensively than a human could. But that same autonomy is what makes people hesitant. Consumers appear willing to experiment with AI as a tool—a research assistant, a recommender, a filter through which to view options. They are far less willing to cede control entirely. And in Singapore's case, they are signaling that they want government to establish guardrails before they do.

The path forward for companies developing these systems is now clearer. Autonomy alone will not win trust. Neither will convenience. What the data suggests is that adoption will depend on building in checkpoints—moments where the human user retains visibility and control—and on establishing credible safety standards that governments can certify and consumers can rely on. Singapore's strong preference for official oversight may also reflect broader regulatory expectations in the city-state, where government involvement in technology governance is already established. For now, the market remains in the early phase: curious enough to try, cautious enough to watch, and organized enough to demand that someone be watching too.