TRM Labs examined $52.7 million across 198.9 million settlements on Coinbase’s x402 protocol since May 2025. After filtering out self-payments and anomalous flows to isolate $25.62 million in likely commerce, the firm estimated that AI agents generated only 0.6%–7.5% of the value. The report notes that ordinary scripts can mimic agent behavior by repeating single-service payments, making it difficult to distinguish true autonomous activity from automated tasks without conclusive modeling.
The study highlights significant gaps in accountability, as on-chain agent registries remain voluntary and underutilized. Despite low current usage, major entities continue investing in infrastructure; Binance launched an x402 payment layer with Agent OS, while Amazon introduced AgentCore Payments with Coinbase and Stripe. USDC accounted for 99.6% of settled value, underscoring stablecoin dominance in this emerging market structure.
This data suggests that the narrative around immediate, widespread agentic commerce is premature. While the technical rails function effectively, the distinction between simple automation and complex AI-driven decision-making remains blurred. The low percentage of verified agent activity indicates that most current volume stems from scripted interactions rather than autonomous economic behavior, highlighting a gap between infrastructure readiness and actual adoption maturity.
From a compliance perspective, the lack of mandatory registration poses operational risks. Without accurate attribution, monitoring systems cannot effectively distinguish between legitimate micro-transactions and potential money laundering or abuse via scripts. The call for "agentic compliance" underscores the need for regulatory frameworks that address high-volume, low-value transactions specifically, moving beyond traditional value-based surveillance methods.
