Newsroom
18 November, 2026 / News / AI / Tags: hyundai, kinexys, agent, card, jpmorgan

Hyundai Card and JPMorgan’s Kinexys platform finished a proof-of-concept in which AI agents managed simulated cash balances, compared foreign-exchange quotes and executed transfers within strict predefined limits
Hyundai Card, the financial services unit of Hyundai Motor Group, and Kinexys by J.P. Morgan have completed a proof-of-concept demonstrating that artificial-intelligence agents can handle corporate treasury tasks without direct human input for every step. The exercise, announced on November 18, 2026, involved only simulated transactions on JPMorgan’s private blockchain platform and moved no real funds.
In the test scenario, two hypothetical overseas subsidiaries maintained deposit accounts. When the balance at one subsidiary fell below a predetermined threshold, the Hyundai Card AI agent monitored the positions, compared foreign-exchange rates offered by multiple banks, and instructed a Kinexys AI agent to convert currency and transfer the shortfall so the receiving account stayed above the required level.
Kinexys developed both the Hyundai Card agent and its own agent, along with the systems that verified identity, authorization, control and validation. Hyundai Card designed the transaction scenario and the authorization framework that set the boundaries for the agents.
Upon receiving an instruction, the Kinexys agent first confirmed the identity and authority of the requesting agent. Transactions that fell inside the approved parameters executed automatically. Requests that exceeded those parameters were either escalated for human review or rejected outright.
To illustrate the approach, Toh cited a hypothetical limit of $5 million under which an agent would be prevented from completing a larger trade. That figure served only as an example and was not the actual threshold applied in the Hyundai Card trial.
The trial covered three core treasury functions: continuous monitoring of account balances, comparison of foreign-exchange quotes across banks, and autonomous cross-border fund movement after conversion. Participants described the exercise as the first instance in which AI agents from separate institutions securely exchanged information and completed financial transactions with each other on a private blockchain.
Kinexys operates as a bank-led, permissioned blockchain designed for institutional payments, tokenized deposits and continuous settlement. Its deposit accounts support multiple currencies, including the euro, and provide near real-time visibility within JPMorgan’s regulated banking framework.
The companies have set a target of the first half of 2027 for a more complex phase that will examine interactions among multiple agents. No commercial launch date has been announced. Hyundai Card has indicated interest in expanding cooperation with JPMorgan into areas such as in-vehicle payments, though those plans remain separate from the current treasury test.
The AI-agent trial follows Hyundai Card’s earlier pilot that moved a $20,000 Tether transfer between subsidiaries on the Avalanche blockchain in about seven minutes. Unlike that earlier exercise, the Kinexys test remained fully simulated and focused on rule-based controls rather than live settlement.
Officials involved noted that technical integration proved relatively straightforward while establishing clear accountability and governance frameworks presented the greater challenge. The proof-of-concept kept all agent activity inside human-defined boundaries, ensuring that autonomous execution occurred only within tightly constrained parameters.









