As commerce transitions into the age of autonomous systems, the most critical battleground is no longer the user interface but the underlying protocol layer that governs how transactions are executed. Insights from Airwallex highlight that this invisible infrastructure will ultimately determine how trillions in global commerce are routed, monetized, and controlled.
While platforms like ChatGPT or Gemini dominate attention as discovery interfaces, the real shift is happening beneath the surface. The key question is no longer how users shop, but how transactions are structured, validated, and settled when machines not humans become the primary actors.
From human checkout to machine-to-machine commerce
The traditional “search, click, buy” model is rapidly being replaced by a fragmented ecosystem where AI agents handle discovery, decision-making, and execution. In this environment, the role of the merchant fundamentally changes. Two distinct layers are emerging. The discovery layer determines how products are surfaced by AI systems, while the protocol layer defines how payments are processed and who retains control over the transaction. As Airwallex outlines, this distinction is becoming critical for businesses navigating the next phase of digital commerce.
At the center of this shift is a growing divide between two competing models. On one side are closed, platform-driven ecosystems such as the Agentic Commerce Protocol (ACP), developed by players like OpenAI and Stripe. These systems aim to keep the entire transaction discovery, checkout, and payment within a single AI-controlled environment. While this approach optimizes speed and conversion, it reduces the merchant’s role to a fulfillment layer, limiting access to customer relationships and data.
On the other side are open, merchant-driven frameworks such as the Universal Commerce Protocol (UCP), supported by companies like Google and Shopify. Rather than centralizing control, these systems allow merchants to publish machine-readable catalogs that AI agents can query. Transactions still occur within the merchant’s own infrastructure, preserving ownership of customer data, checkout experience, and loyalty mechanisms.
The rise of machine-native payments and identity layers
Beyond traditional retail transactions, a new category of commerce is emerging machine-native payments. In this model, software systems transact directly with one another, enabling autonomous purchasing and resource allocation. Protocols such as MPP and x402 are designed to support these interactions, allowing AI agents to execute payments across multiple rails, including cards, stablecoins, and microtransactions. This introduces entirely new use cases, from automated service procurement to real-time API monetization. At the same time, identity and authorization are becoming central challenges. As transactions shift from human intent to machine execution, verifying legitimacy becomes critical. Initiatives such as “Know Your Agent” frameworks are being explored to ensure that automated systems act within defined permissions and do not introduce systemic risk.
For merchants, this evolving landscape creates both opportunity and uncertainty. Locking into a single protocol too early could limit flexibility in a rapidly changing environment. Instead, the focus is shifting toward building infrastructure that is adaptable, interoperable, and capable of handling multiple transaction standards. This includes structuring product data for machine consumption, maintaining payment stack flexibility, and ensuring compliance with emerging regulatory expectations around AI-driven transactions. At a broader level, the rise of agentic commerce introduces a potential tension between efficiency and control. While automation promises frictionless transactions, it also risks reducing the direct relationship between brands and consumers. This could lead to a countertrend, where certain businesses differentiate through human-centric experiences that cannot be replicated by automated systems.
As the industry moves forward, one conclusion becomes increasingly clear: the future of commerce will not be defined by who owns the interface, but by who controls the protocol layer that powers it.
