CFOtech Asia - Technology news for CFOs & financial decision-makers
Asia
Hyundai Card, J.P. Morgan test AI-led transactions

Hyundai Card, J.P. Morgan test AI-led transactions

Wed, 18th Nov 2026 (Today)
Karen Joy Bacudo
KAREN JOY BACUDO Finance Editor

Hyundai Card has completed a proof of concept with Kinexys by J.P. Morgan for AI agent-led financial transactions. The test examined whether AI agents at separate financial institutions could communicate and carry out transactions in a controlled setting.

The work covered cash management, foreign exchange and fund transfers. Hyundai Card designed the transaction scenario and set the authorisation and control framework, while Kinexys built the AI agents and implemented identity verification, authorisation, control and validation measures.

The project used a simulated arrangement between hypothetical overseas subsidiaries. In the scenario, one subsidiary sent funds to another when the receiving entity's account balance fell below a preset threshold.

An AI agent assigned to Hyundai Card monitored account balances, compared foreign exchange quotes from multiple banks and sent instructions for the required currency conversion and transfer. A Kinexys AI agent then checked the identity and authority of the Hyundai Card agent before acting on those instructions.

Transactions within agreed limits were processed automatically. Any transaction outside those limits was either referred to a human for approval or rejected.

Controlled test

The companies described the exercise as a demonstration of how AI agents could exchange information securely and execute transactions in a controlled environment. The model went beyond software that only analyses cash positions or follows fixed instructions by testing whether agents from separate institutions could communicate and make decisions within set controls.

Hyundai Card said the work is part of its effort to examine payment technologies emerging in international financial markets. Kinexys, J.P. Morgan's blockchain unit, has been developing programmable payments and related infrastructure for institutional use.

"The successful completion of this PoC with Kinexys by J.P. Morgan marks a significant step toward a next-generation financial transaction model in which AI agents can autonomously manage corporate funds and execute foreign exchange transactions with minimal human intervention," a Hyundai Card spokesperson said.

"Building on our extensive expertise and competitive capabilities in payments, we plan to expand our collaboration with J.P. Morgan into new areas, including in-vehicle payments," the spokesperson added.

Wider push

The initiative follows another Hyundai Card payments test involving stablecoin-based cross-border remittances. In that earlier pilot, run with Tether and Avalanche, the company completed a transfer of funds between Hyundai Motor subsidiaries in the United States and Mexico in about seven minutes, including transfer and verification.

That trial focused on cross-border payments using blockchain-based infrastructure, while the latest proof of concept examined how AI agents might handle treasury-related tasks and transaction execution between institutions. Together, the projects reflect Hyundai Card's interest in applying artificial intelligence and digital asset-based systems to payment operations.

Hyundai Card, the financial arm of Hyundai Motor Group, is based in Seoul. Beyond consumer card products, it has pushed into data science, proprietary technology and cross-border payment testing as it seeks a larger role in financial technology.

For J.P. Morgan, the test adds to a broader push to use blockchain and programmable transaction systems in mainstream finance. Kinexys has focused on institutional applications where automated controls, identity checks and settlement processes can be built into transaction flows.

The proof of concept did not involve live customer transactions. Instead, it was designed to test whether AI agents could operate under defined controls, with routine transactions handled automatically and human intervention required for items outside approved limits.

That distinction is likely to matter as banks and payments groups assess where artificial intelligence can be allowed to act directly in financial operations. Treasury functions such as liquidity management, foreign exchange execution and internal fund transfers are areas where repetitive tasks, clear thresholds and formal approval chains can make controlled automation easier to test.

The latest work also points to a question facing financial institutions: how to govern machine-to-machine communication when instructions can trigger the movement of money. In Hyundai Card's test, that issue was addressed through identity verification, permission checks and validation rules before any transaction could proceed.

The simulated set-up was designed to keep the receiving subsidiary's account above a required balance by having the sending subsidiary convert and transfer only the amount needed to cover the shortfall.