Image generated by Deeptech Times using ChatGPT
At this year’s SuperAI conference in Singapore, much of the conversation around AI revolved around capability. AI agents are becoming more autonomous, more sophisticated and increasingly capable of executing tasks that once required human intervention.
Yet, for Dr. Bo Bai, CEO, chairman and co-founder of MetaComp, the defining challenge facing the financial industry is whether AI agents can become trustworthy, not smarter.
As the financial sector moves towards autonomous AI agents capable of making decisions and executing transactions, Bai believes the industry is approaching a pivotal inflection point. The future of finance, he argues, will not be defined solely by intelligence, but by governance.
“The intelligence level is already sufficient,” he says. “The challenge now is trustworthiness.”
The rise of agentic finance
The emergence of agentic AI coincides with another transformation that Bai has been championing for years: the convergence of TradFi and digital assets.
MetaComp has been at the forefront of enabling hybrid financial models that combine fiat currencies, stablecoins and tokenised assets. What was once considered experimental is increasingly becoming operational reality.
According to Bai, businesses across APAC are increasingly accepting stablecoins for cross-border trade, particularly when exporting products to markets across the Middle East, Africa and Latin America. While buyers may prefer paying in digital assets, suppliers and manufacturers still largely operate in traditional currencies.
This has created growing demand for hybrid payment infrastructures that seamlessly bridge both worlds.
“The flow of taking in stablecoins, converting them into fiat and then making payments has been growing substantially,” says Bai.
The trend is reflected in MetaComp’s own growth. The company now processes approximately US$1.2 billion in monthly transaction volume across these payment corridors.
At the same time, new forms of tokenised value are beginning to emerge. Beyond stablecoins, Bai sees growing interest in tokenised gold, particularly among businesses and investors in Southeast Asia and the Middle East who view gold as a trusted store of value during periods of geopolitical uncertainty and currency volatility.
Taken together, these developments point towards a future where financial systems become increasingly programmable, interconnected and autonomous. And that is precisely where AI agents enter the picture.
From people-to-people to agent-to-agent
Historically, financial services were built around interactions between people and institutions. Digital transformation introduced interactions between people and platforms. AI may introduce something fundamentally different.
“We are moving towards a world of people-to-agent and agent-to-agent interactions,” says Bai.
In this emerging model, AI agents may not simply assist users with recommendations. They could execute payments, manage wallets, monitor transactions, perform compliance checks and make financial decisions on behalf of individuals or organisations.
To support that future, Bai believes the industry needs a governance framework equivalent to the KYC and KYT standards that underpin modern financial compliance.
That thinking led MetaComp to introduce its “Know Your Agent” (KYA) framework.
The concept is deceptively simple: before a financial institution allows an AI agent to access financial services, it should understand who owns the agent, who controls it and how it makes decisions.
“If financial services are increasingly provided to agents, then we need to know which agents are allowed to provide services and which agents are allowed to receive them,” Bai explains.
Inspired in part by Singapore’s leadership in AI governance, KYA seeks to establish accountability mechanisms for AI-native financial systems before regulation catches up.
Accountability cannot be outsourced
One of the biggest misconceptions surrounding autonomous AI is the notion that responsibility can somehow be delegated to the technology itself. Bai firmly rejects that idea.
When an AI-generated marketing campaign fails, the consequences may be limited. When an autonomous financial agent makes an error involving money, the consequences can be far more severe.
“There is a very thin margin for error in financial services,” he says. “If money disappears, you cannot simply create another version.”
For that reason, Bai argues that human oversight remains essential, regardless of how advanced AI becomes.
Under the KYA framework, accountability ultimately rests with the legal entity or individual that owns and controls the agent. Financial institutions must understand who is behind the AI, verify their identity and ensure sufficient oversight exists over the agent’s activities.
In other words, governance cannot be separated from deployment. It must be embedded into the architecture itself.
The production challenge
While many organisations are experimenting with AI agents, Bai draws a clear distinction between proof-of-concept demonstrations and production-grade deployments. The difference comes down to risk management.
“We still don’t know what we don’t know,” he admits.
Technologies such as agent orchestration, autonomous decision making and Model Context Protocol (MCP)-based interactions remain relatively new. As a result, organisations cannot assume that successful pilot projects automatically translate into enterprise readiness.
Instead, Bai advocates a measured approach: identify risks, govern them carefully, deploy gradually, collect operational data and continuously refine controls.
This philosophy mirrors how financial regulation itself has evolved over time. New technologies often emerge faster than regulatory frameworks can adapt. Industry participants therefore have a responsibility to experiment responsibly while providing real-world evidence that regulators can learn from.
Singapore’s regulatory approach to digital assets offers a useful precedent.
Bai believes AI governance may follow a similar path, with Singapore potentially emerging as one of the first jurisdictions to establish comprehensive frameworks for agentic financial services.
Beyond APIs: The next connectivity revolution
Another development Bai believes could accelerate the adoption of agentic finance is the emergence of MCP and AI-native skills frameworks. He compares the shift to the rise of APIs during the open banking era.
For years, APIs have served as the backbone of financial connectivity, enabling banks, platforms and enterprises to exchange information and services. MCP introduces a new paradigm where AI agents can interact with systems more naturally and efficiently.
“The concept of open banking remains,” says Bai. “But we are using a far more advanced way to create connectivity.”
The result could be dramatically shorter integration cycles and significantly faster deployment of financial services within AI ecosystems. For AI-native companies building agentic commerce, agentic marketing or autonomous business workflows, these capabilities may become foundational infrastructure.
Building the future responsibly
The excitement surrounding agentic AI often focuses on what machines will be capable of doing in the future. Bai’s perspective offers an important counterbalance.
History shows that transformative technologies are frequently adopted first by those seeking to exploit them. The same risks that accompanied the rise of the internet, blockchain and digital payments will inevitably accompany autonomous AI systems. That reality makes governance not a constraint on innovation, but an enabler of it.
The future of finance may indeed be autonomous. Payments, compliance, wealth management and financial operations may increasingly be executed by AI agents operating at machine speed.
But if that future is to scale responsibly, the industry must answer a fundamental question first. For Bai, the answer begins with governance. And in the age of agentic finance, trust may ultimately prove to be the most valuable innovation of all.
