[TokenPost Column] The Prelude of the 'Agentic Economy': The Day AI Becomes the Bank Vault Manager

Now is the era of the “Agentic Economy.” This term, selected as a core keyword by the January 2026 issue of the premium monthly publication “Blockchain Business Review” under TokenPost, signifies a new economic paradigm where AI transcends being a mere tool to become an autonomous economic actor. Moreover, this massive trend is also reflected in the most conservative and core sector of the financial system—“payment and settlement” systems.

Recently, a working paper released by the Bank for International Settlements (BIS) and the research team of the Bank of Canada demonstrated that generative AI (Gen AI) has the potential to assume the role of a bank’s “Cash Manager,” implying that the agentic economy is not a distant sci-fi future but an imminent reality.

◇ BIS Experiment: AI Unlocks the “Liquidity Management” Challenge

The BIS research team conducted an experiment where liquidity management authority in the wholesale payment system was granted to a general-purpose generative AI model. The task assigned to this AI agent was far from simple. It needed to prepare for liquidity shocks amid uncertain cash flows and delicately balance the trade-off between settlement delay costs and liquidity holding costs.

Surprisingly, even a general AI model without specialized financial training made “precautionary” decisions similar to those of skilled human managers.

The AI agent made judgments to delay current small payments to prepare for future large emergency payments;

It recalibrated priorities by calculating the likelihood of fund inflows;

And derived an optimal strategy that minimizes settlement delays while preserving liquidity.

This strongly proves that AI has evolved beyond simply executing input commands to become an “economic intelligent agent” capable of calculating risks and taking actions under uncertainty.

◇ “Faster and More Flexible than Reinforcement Learning (RL)”… Innovations Enabled by Gen AI

The message conveyed to the Korean financial sector by this research is very clear: “Speed” and “Efficiency.”

Previously, reinforcement learning (RL), the mainstream model for financial optimization, was like teaching children payment and settlement rules one by one from scratch. To learn the optimal strategy, it required tens of thousands of simulations, massive data, and a long time.

In contrast, AI agents based on generative AI possess “zero-shot” or “few-shot” learning capabilities. They can understand contexts and perform real-time reasoning with just prompts, eliminating the need for extensive long-term training to engage in complex financial operations. This means financial institutions can significantly reduce operational costs and accelerate prototyping by introducing AI agents.

[BBR Vol. 16] “AI Has Wallets and Trades Autonomously”… The New Financial Order Brought by the 2026 “Agentic Economy”

◇ The Future of Finance: “AI Sandboxes” and the Role of Humans

Of course, full authority over bank vaults cannot be handed to AI immediately. The BIS report emphasizes the “Human-in-the-loop” model and highlights the importance of human managers who perform the final validation of AI decisions. At the same time, it warns of the risks of AI relying on past data and making misjudgments in unforeseen “black swan” scenarios.

But the direction is set. Policymakers can leverage these AI agents to build “multi-agent payment settlement simulators.” They can safely test systemic deadlocks or resilience in crisis scenarios within virtual sandboxes and experiment with new regulatory policies.

◇ The Attitude We Should Have to Welcome the ‘Agentic Economy’

The reason why TokenPost’s January issue focuses on the ‘agentic economy’ is very clear. 2026 will become the inaugural year when AI upgrades from being a human auxiliary tool (Tool) to an autonomous agent (Agent).

The Korean financial market has accelerated its AI transformation through the introduction of AI bankers and active robot advisory services. BIS’s research further demonstrates that AI can go beyond simple responses to become a core business entity in areas like fund management.

Domestic financial institutions should now speed up the adoption of purpose-fit AI solutions to improve operational efficiency, while also preparing thoroughly for risk assessment and regulatory compliance. BIS’s research is just the beginning of this massive transformation. The future of finance now depends on “who can employ smarter intelligences.”