0x5c116e685c11…5c116e65
Hyundai Card tests AI treasury handoff on Kinexys blockchain
Hyundai Card and JPMorgan’s Kinexys ran a fully simulated AI-to-AI currency conversion on November 18, 2026, with no real funds moved and authorization limits enforced.

Outputs
The simulated AI-agent treasury transaction took place on November 18, 2026.
Hyundai Card’s agent compared bank foreign exchange quotes and instructed the Kinexys agent to convert currency and transfer funds.
No real money moved; transactions above preset authorization limits were escalated for review or rejected.
Kinexys has processed more than $4 trillion cumulatively, averaging approximately $7 billion in daily volume.
Expanded multi-agent interactions and negotiation capabilities are scheduled for the first half of 2027.
Nathan Brooks
Hyundai Card and JPMorgan’s Kinexys completed a fully simulated treasury transaction on November 18, 2026, in which an AI agent compared foreign exchange quotes, selected a route and instructed a bank’s AI agent to convert currency and transfer funds. No human typed the order, and no real money moved.
The proof-of-concept ran on blockchain deposit accounts operated by Kinexys, JPMorgan’s private blockchain platform. The test was described as the “first secure handoff of financial transactions between institutions by AI on a private blockchain,” according to the project description.
How the simulated handoff worked
Two AI agents performed the exercise. Hyundai Card’s agent played a corporate treasurer and monitored balances for hypothetical overseas subsidiaries, reflecting the cash positions a multinational finance team checks each day. After identifying a need, the agent obtained foreign exchange quotes from multiple banks and compared them.
The Hyundai Card agent then chose a route and sent instructions to the Kinexys AI agent. The receiving agent handled the currency conversion and fund transfer inside the simulation. The test therefore connected one institution’s agent directly with another’s operational workflow, without a person entering the payment instruction.
Preset authorization limits supplied the principal control. Any transaction exceeding those limits either went to a person for review or was rejected outright. Executives involved described the trial as “a meaningful step in exploring AI-driven automation for treasury functions while keeping governance and compliance intact.”
Why the private ledger matters
Kinexys operates at institutional scale rather than as a laboratory experiment, according to the source. Its processing record supports that distinction:
- Cumulative processing: more than $4 trillion in transactions since inception.
- Average daily volume: approximately $7 billion, equivalent to about $2.56 trillion over 365 days at the same pace.
The platform supports programmable instructions and near-instant transfers outside traditional banking hours. It runs a private ledger for bank deposit accounts, not a public network with open participation. That architecture places the agent workflow inside a bank deposit-account environment, rather than an open public network.
For banks and corporates, the distinction between a simulated handoff and a live deployment remains material. The exercise demonstrates a workflow and its control points, but the description provides neither general production availability nor a customer rollout schedule.
What does this change for treasury teams?
For treasury teams, the immediate change is an operating pattern for automating multi-bank currency decisions and transfers, not a new funding channel. The simulation shows where an agent can act within an approved envelope and where human review remains necessary when an instruction exceeds that envelope.
The next phase moves beyond a fixed sequence of roles. Agents may negotiate with one another directly rather than only compare quotes and select a route. That shifts the operational focus from executing a known workflow to defining the rules under which independent software agents bargain on a corporate’s behalf.
The next scheduled checkpoint is the first half of 2027, when Hyundai Card and Kinexys plan to expand multi-agent interactions and add negotiation capabilities.
via Crypto Briefing (Source)
More from Nathan Brooks
Show full bio
Market editor covering money, funding and market shifts around crypto and web3 industry at Mempool Brief.
1,512 articles