trending
The global race to define how Agentic AI will handle money reached a milestone this week as Google unveiled its Agent Payments Protocol (AP2).
Google Agentic AI protocol
The open standard, created in partnership with more than 60 payments and technology companies, is designed to provide a common set of rules for how autonomous AI agents can initiate and complete transactions.
Developed with input from Adyen, American Express, Ant International, Coinbase, Mastercard, PayPal, Worldpay and others, the protocol is intentionally payment-agnostic.
It supports a broad spectrum of settlement types, from debit and credit cards to real-time transfers and even stablecoins.
Agentic AI
For Google, the goal is to create a single framework through which agentic AI systems – digital assistants that act independently on behalf of a user – can interact with merchants and financial institutions in a secure and predictable manner.
Agentic AI represents a major departure from existing consumer payments.
Instead of a customer pressing “pay”, an AI assistant might be authorised to book flights, order groceries, or purchase concert tickets without human intervention.
Autonomy Introduces Risk
That autonomy, however, introduces new risks: verifying that the agent is genuinely authorised, preventing fraud, and ensuring the payment is both secure and accountable.
AP2 seeks to resolve these challenges through the use of cryptographically-signed Mandates.
In practice, these are tamper-proof digital instructions that prove both a user’s intent and the boundaries of an agent’s authority.
For a one-off purchase, the user issues an “Intent Mandate” (such as “buy me a gold watch”) which is later confirmed by a “Cart Mandate” specifying the exact item and price.
For more open-ended tasks – like securing tickets the moment they are released – the agent collects additional parameters on timing and pricing before finalising the purchase.
Supporters argue this framework could provide banks and card networks with the assurances they need to treat agent-initiated transactions with the same level of trust as a traditional card swipe.
“Google’s new protocol provides a common language to delegate credentials, set boundaries, and create accountability,” said Rodri Touza, co-founder of Crossmint, one of the protocol’s contributors.
Customer Engagement
Merchants, meanwhile, are eyeing the technology as a new channel for customer engagement.
AI agents could be used to recognise loyal shoppers, provide early access to limited stock, or even conduct agent-to-agent negotiations on pricing.
Yet risks remain: malicious bots might exploit loyalty accounts or simulate demand to distort inventories, meaning fraud detection will need to evolve well beyond identity verification alone.
For the financial sector, the emergence of AP2 is a pivotal step in building the guardrails around agentic commerce.
It offers a blueprint for interoperability at scale – but also signals the beginning of a much broader debate about how to balance automation, trust and security in the age of AI-driven payments.
Comments