Agentic Commerce Is Coming. Most Payment Rails Aren't Ready For It

WCT ·

Right now, every payment still assumes a human is in the loop: clicking, confirming, typing in a card number, hitting "pay." That assumption is about to break. AI agents are starting to handle purchases on their own. A user sets preferences once, spends up to $500 on hotels, always settles in stablecoins, never books anything non-refundable, and the agent takes it from there. No checkout page. No confirmation screen. The transaction just happens in the background, and the human only finds out about it after the fact, if they check at all. That shift has a name: agentic commerce. And it's forcing a hard question for every PSP, merchant, and payments team: can your current stack actually process a payment nobody clicked "pay" on? Agentic commerce is what happens when an AI agent, not a person, initiates and completes a transaction on a user's behalf. The agent might be booking a flight, restocking inventory, subscribing to a tool, or paying another agent for a service. In every case, the pattern is the same: no UI, no manual approval step, just an API call that needs to clear compliance, settle funds, and confirm the transaction, all without a human watching. This isn't a distant, theoretical future. Merchants are going to have customers running AI shopping agents within the next 12 to 24 months. The infrastructure question is whether their payment rail can keep up, or whether it was built for a world where a person is always the one pressing the button. It's worth being precise about what's actually new here. E-commerce has already automated the storefront. Subscriptions already automate the billing cycle. What agentic commerce automates is the decision itself, the moment where a human used to look at a price, compare it against a budget, and choose to proceed. That decision now happens inside the agent's logic, and the payment rail underneath it has to be fast and trustworthy enough that no one needs to double-check it in real time. Card networks were designed around a physical terminal and a human approving each step. That architecture has served the industry well for decades, and it's adapting fast, but it wasn't built with autonomous, API-first transactions in mind. Every layer, from authorization to fraud checks to settlement, still assumes a person is somewhere in that chain, even if that person is just clicking "confirm" on a phone. Fraud models are a good example of where this shows up. Card networks flag unusual purchase patterns based on how humans normally behave: location, timing, device, spending history. An AI agent making dozens of small, scheduled purchases across different merchants doesn't look like a human at all. It looks like exactly the kind of activity those models were built to catch. That's not a small compliance footnote; it's a structural mismatch between how card rails detect risk and how agents actually transact. Stablecoin rails evolved differently. They were API-first from day one, which means they don't need to be retrofitted for agents. They were simply built for exactly this kind of transaction from the start, where an API call, a signature, and a settlement confirmation are the entire interaction. There's no terminal to simulate, no human-shaped fraud model to work around. Now exploring @Google 's agentic commerce protocol: AP2 (part of UCP) and buying things using @WalletConnect Pay. This time using Gemini CLI as my agent and asking to buy stuff. Flow works very similar to my previous MPP example — our sample e-com app stays as a Checkout Mandate… https://t.co/RXhD2H5pPq pic.twitter.com/Duqu0iUmet No UI required. Every payment action is accessible through an API. An agent can initiate, authorize, and settle a transaction without ever touching a human-facing interface. There's no checkout page to render, no confirmation modal to click through, and nothing that assumes a screen is even involved. Compliance runs automatically. Sanctions screening and Travel Rule checks execute inside the transaction itself. The agent doesn't need to understand compliance rules; the infrastructure below it handles that automatically, before value ever moves. That matters because an agent can't call a compliance officer if something looks unusual, the check has to happen inline, every time, without exception. It works across any chain. The agent doesn't need to know which blockchain a user's funds sit on. Routing and conversion happen automatically underneath, so the agent experience stays simple no matter how complex the settlement layer gets. This also future-proofs the integration: as new chains and assets get added to the network, the agent doesn't need to be reprogrammed to use them. A few patterns are emerging faster than most teams expect: Travel and booking agents. A user sets a budget and preferences once; the agent books flights, hotels, and transport, paying instantly in stablecoins without a checkout flow. No more comparing five tabs and re-entering card details on each one. Procurement and subscription bots. Business agents renew software licenses, restock supplies, or pay vendors automatically based on preset rules and spend limits, cutting out the manual approval queue that slows down routine purchasing. Personal shopping agents. Consumers delegate recurring or routine purchases, groceries, refills, gifts, to an agent that executes the purchase the moment conditions are met, like a price drop or a restock alert. Agent-to-agent micropayments. As agents start paying other agents for data, compute, or services, transaction volumes will spike in ways that make manual checkout completely impractical. Think an agent paying a few cents per API call to another agent, thousands of times a day. Dynamic pricing and negotiation. Some agents won't just execute a fixed purchase, they'll negotiate within a range and settle instantly once terms are agreed, something that's effectively impossible with a payment rail that assumes a fixed checkout total and a human confirming it. Every one of these needs the same thing underneath: a payment rail that settles in seconds, runs compliance without a human in the loop, and doesn't care which chain or wallet is involved. WalletConnect already connects 500M+ users across 700+ wallet providers, with $400B+ in volume moving through the network in 2025 alone. That scale matters here because agentic commerce doesn't just need speed; it needs reach. An agent transacting on a user's behalf can't be limited to one wallet type or one chain; it needs the same universal reach a human already expects when they tap to pay. WalletConnect Pay sits on top of that connectivity layer specifically so PSPs don't have to build agent-ready infrastructure from scratch. Compliance, sanctions screening, and Travel Rule checks are already embedded at the protocol level, which means an agent-initiated transaction gets the same protection as a human-initiated one, automatically. There's no separate "agent mode" to build and maintain, the same rail that handles a person tapping to pay at a coffee shop handles an agent settling a subscription renewal at 3am. For PSPs specifically, that means the agentic commerce wave doesn't require a second integration effort down the line. The API-first design means the work done to support stablecoin acceptance today is the same work that supports agent-initiated payments tomorrow. The teams that integrate agent-ready, stablecoin-native payment rails now won't be scrambling when this wave actually hits. The ones who wait will be retrofitting infrastructure that was never designed for autonomous transactions in the first place, while their competitors are already live and processing agent-initiated volume at scale. Agentic commerce isn't a future trend to monitor from the sidelines. It's an infrastructure decision that's already due, and the gap between the teams who prepare for it and the teams who don't is going to show up first in who can actually process the transaction, not in who has the better marketing story about AI.

DYAX Investor Sentiment

Bullish (Long) 33% · Bearish (Short) 67%

380 participants

Related News

원문 보기 — WCT