This effort has shown the potential for artificial intelligence (AI) agents to “autonomously and securely execute financial transactions,” Hyundai Card said in a news release provided to PYMNTS Wednesday (Nov. 18).
The proof of concept (PoC) completed by the two companies was designed to “validate the feasibility of automating financial operations, including cash management, foreign exchange (FX) and fund transfers, through communication between AI agents,” the release added.
Hyundai said the PoC centered around simulated transactions between hypothetical overseas subsidiaries using AI agents to handle and transfer funds.
Hyundai Card designed the transaction scenario and created the authorization and control framework for the AI agents. Using this framework, Kinexys developed the AI agents for both companies and implemented identity verification, authorization, control and validation mechanisms to secure transactions.
“The PoC represents a first-of-its-kind demonstration of how AI agents can securely exchange information and execute financial transactions within a controlled environment,” the release said. “While conventional AI-powered financial automation has primarily focused on analyzing cash positions or executing transactions based on predefined rules, this PoC explored a more advanced model in which AI agents from different institutions communicate, make independent decisions and carry out financial transactions autonomously.”
In other agentic AI news, PYMNTS wrote recently about a scenario in which a portal might reject an agentic payment. The report described a restaurant paying its food distributor, packaging supplier and equipment service provider.
“An agent could submit all three approved payments if it has access to the suppliers’ systems and each payment meets the applicable rules,” the report said. “If the packaging supplier’s portal rejects the card, the bookkeeper may need to arrange another way to pay. If the equipment company’s portal shows a higher balance than the approved invoice, someone has to check whether the supplier revised the bill or whether the portal includes another charge.”
A supplier may also grant a discount in exchange for getting paid earlier. Tungsten Automation CEO Peter Hantman told PYMNTS that the gap between invoice approval and payment is an opportunity to examine such offers.
Research by PYMNTS Intelligence shows that fewer bills are being paid ahead of schedule among larger businesses, with the share of payables paid early dropping from 37% to 17% between editions of the Working Capital Index.
“For a buyer paying on the due date, a portal rejection leaves less time to correct the problem,” the report said. “If two of the restaurant’s three payments go through, the bookkeeper needs to know which supplier is still waiting and whether the third payment failed because of the card, the invoice or the portal.”
For all PYMNTS AI and digital transformation coverage, subscribe to the daily AI and Digital Transformation Newsletters.