JPMorgan and Hyundai Card advance AI agent transaction testing in corporate treasury
Banks and payment groups are exploring how artificial intelligence agents can take on more financial tasks as automation moves deeper into treasury and payments. JPMorgan and Hyundai Card have completed a simulated foreign exchange test without direct human intervention, with broader multi-agent trials targeted for the first half of 2027.
Highlights
- JPMorgan and Hyundai Card completed a proof-of-concept where AI agents executed a simulated foreign exchange transaction on the Kinexys blockchain without human involvement.
- In the first half of 2027, the next testing phase will introduce multiple autonomous agents interacting within a shared treasury transaction environment, extending agent capabilities beyond simple execution.
- JPMorgan highlights that the main commercialization challenge is governance, with Kinexys implementing frameworks to confine agent actions to pre-approved limits to mitigate operational and control risks.
Next phase targets multi-agent treasury use
As first reported by CNBC, JPMorgan and Hyundai Card have completed a proof-of-concept in which AI agents carried out a simulated foreign exchange transaction without direct human involvement. The project is now moving toward a more complex testing phase in the first half of 2027, when multiple autonomous agents are expected to interact within the same transaction environment.The initial trial involved a Hyundai Card AI agent requesting a foreign exchange transaction to maintain euro liquidity in its Kinexys blockchain deposit account. JPMorgan said the test was the first private blockchain-based autonomous bank-agent-to-client-agent trial for corporate treasury management.
Wee Kee Toh, global head of Kinexys Insights at Kinexys by JP Morgan, said the next stage is to examine how autonomous agents behave when they move beyond simple execution and begin interacting with one another in an ecosystem. That would extend the role of AI agents from carrying out instructions to potentially negotiating and completing payment-related actions across financial workflows.
Governance remains the main commercialization hurdle
JPMorgan has not set a timeline for commercial deployment of the technology, even though the initial proof-of-concept has been completed. The bank says the larger challenge is not building the system itself, but creating governance, internal controls and transaction limits that would make autonomous agent activity acceptable for real-world financial use.To address that issue, Kinexys has developed a framework that confines AI agents to predefined transaction parameters and automatically rejects requests outside authorized limits. That approach is aimed at reducing operational and control risks as financial institutions assess whether AI agents can be trusted with a wider range of treasury and payment functions.
As we previously wrote, the EBRD’s Transition Report spotlighted artificial intelligence and defence spending as emerging competitiveness drivers.