An AI agent completed a full purchase and payment on its own for the first time on BLIK, Poland’s dominant mobile payment scheme, without a human present at the moment of checkout. Working inside an AI shopping assistant, the agent monitored the online store Your KAYA for a specific hand cream to come back in stock, then bought it the moment it did, confirming the transaction with a six-digit BLIK code the user’s banking app had already been authorized to release.
The mechanics matter as much as the milestone. BLIK, the merchant platform Juo, and payment processor PayU built the pilot so the agent never holds payment credentials itself. It operates inside spending limits and consents the user set in advance, and it cannot trigger a transaction the user has not already authorized. “Agent autonomy does not mean that consumers lose control,” said Katarzyna Prus-Malinowska, Head of Product Management at PayU, describing an architecture where the agent executes within boundaries the human already approved rather than acting as an independent financial actor.
The original insight here is about sequencing, not novelty. Card networks have spent 2026 publishing agent-identity frameworks and interoperability standards for a future state of agentic commerce that mostly has not arrived yet. BLIK’s pilot shows a national payment scheme running a real, if narrow, transaction before the identity plumbing behind it is fully built. That gap, live pilots outrunning the trust infrastructure meant to govern them, is exactly what frameworks like the ones Visa, Mastercard and card networks are racing to formalize are trying to close before the pilots scale into volume.
For a payments leader, the immediate takeaway is that agentic checkout is no longer a roadmap item at the infrastructure layer. It is running, at small scale, on real national rails.
Read more: Agentic Commerce Splits Into Reasoning and Payment Layers and EMVCo Sets Ground Rules for AI Agents Paying by Card.
Source: BLIK
