AI and cryptocurrency payments are increasingly intertwined in discourse, as autonomous agents may require more than mere intelligence to function effectively in commercial settings. They would also require a mechanism to transfer value independently of payment systems primarily designed for human utilisation. For Alex Kozenko, infrastructure stands as the pivotal concern. The essential observation is not that AI agents are currently revolutionising payments on a large scale. Autonomous transactions would impose varying requirements on payment rails. Kozenko posits that the systems employed by AI agents must be programmable, perpetually accessible, and compatible with machine-driven operations. That is where cryptocurrency enters the discussion. Kozenko contended that crypto infrastructure inherently meets those criteria due to its programmability and 24/7 availability.
The open question is whether those features will be sufficient to render crypto payments practical for agentic commerce. A separate study from the Bitcoin Policy Institute, released on March 3, 2026, offers context for the broader conversation, although it was not linked to Kozenko’s remarks. The study evaluated 36 advanced AI models from Anthropic, DeepSeek, Google, MiniMax, OpenAI, and xAI across 9,072 open-ended monetary scenarios. The study found Bitcoin was selected in 48.3% of all responses, more than any other option, while stablecoins followed at 33.2%. Over 90% of responses indicated a preference for digitally native money, such as dollar-pegged stablecoins, in comparison to traditional fiat currencies. According to the study, no model selected fiat as its primary preference.
The results also indicated a division among various applications of capital.Bitcoin dominated store-of-value scenarios at 79.1%, whereas stablecoins excelled in everyday payment scenarios at 53.2%. The study does not establish how genuine AI agents will operate in commercial environments; however, it elucidates the rationale behind the discourse surrounding digital-native currency in conjunction with autonomous transactions. Kozenko stated that agentic payments have not yet become a prevalent commercial reality. His timeline indicates that shift is expected to occur in approximately two to three years. Today’s decisions hold significant weight, as companies may be developing the systems that future AI agents will either utilise effectively or find challenging to navigate.
The term “machine-readable interfaces” highlights an ongoing technical challenge that remains unresolved. Kozenko’s reference suggests that payment systems designed for autonomous agents must be comprehensible and operable by software, rather than solely by humans. In the absence of that layer, while programmable payment rails might be present, AI agents could still find themselves without a viable method to leverage them effectively at scale. The remaining question is not whether AI and crypto payments are garnering significant interest. The inquiry centers on the capability of payment companies to establish infrastructure that transforms autonomous transactions from a mere concept into a functioning commercial system. Until machine-readable interfaces and genuine agentic payment use cases reach maturity, the future Kozenko envisions remains plausible yet unresolved.