Amazon’s Bedrock AgentCore Payments, developed in partnership with Coinbase and Stripe, is revolutionizing the way AI agents interact with paid services. This cutting-edge technology allows AI agents to discover services, authenticate, and make payments using stablecoins and x402, all while adhering to preset spending limits.
As AI agents begin to have their own wallets even before traditional merchants have caught up, shopping agents are already making purchases across the open internet. However, the real challenge arises once the payment is completed. Edgars Nemse, CEO of the GenLayer Foundation, highlighted the issue of determining whether the services or goods were delivered as promised after the payment has been processed.
Nemse emphasized that while the payment process itself is straightforward and deterministic, ensuring that the outcome meets the buyer’s expectations is a more complex judgment call. The table below outlines the different stages of a transaction, highlighting what AI agents can do and the unresolved issues that remain:
1. Discovery: Finding merchants, APIs, content, and services
– Unresolved Issue: Whether merchants trust unknown agents
– Importance: Controls who gets distribution
2. Authorization: Using mandates, credentials, and spending limits
– Unresolved Issue: Whether the agent stayed within the user’s intent
– Importance: Determines who bears liability
3. Payment: Paying with cards, stablecoins, x402, or wallets
– Unresolved Issue: Payment success does not guarantee satisfaction
– Importance: Settlement is deterministic
4. Fulfillment: Receiving goods, services, or API access
– Unresolved Issue: Was the outcome delivered as promised?
– Importance: Requires judgment
5. Dispute: Submitting complaints or refund requests
– Unresolved Issue: Who adjudicates outside a platform?
– Importance: Determines whether open commerce can scale
The integration of AI agents into the payment process has the potential to streamline transactions and reduce friction. However, it also raises concerns about the increase in disputes and the potential impact on merchant costs. Regulators, such as the Reserve Bank of Australia, have started to address these issues and are expected to announce regulatory priorities by the end of 2026.
In order to address the challenges associated with disputes in agentic commerce, innovative solutions are being developed. For example, GenLayer’s AI consensus model aims to provide a neutral dispute resolution process by using validators to reach consensus on outcomes. This approach ensures that decisions are enforced on-chain and are open to appeal, creating a fair and transparent system for resolving disputes.
Overall, the future of agentic commerce hinges on the development of standards that ensure verifiable mandates, merchant evidence records, and efficient dispute resolution mechanisms. By establishing clear guidelines and protocols, AI agents can navigate services and transactions with confidence, paving the way for a new era of seamless and secure online commerce.
