# The Dodo Digest: AI Agents Are Already Spending Money. Are You Ready for That? > AI agents are already making real payments - Mastercard just proved it. We shipped v1.93.0 with business-level payment method control and a rebuilt checkout to handle both AI-driven and human flows. - **Author**: Rishabh Goel - **Published**: 2026-04-03 - **Category**: Newsletter - **URL**: https://dodopayments.com/blogs/newsletter-3april --- **TL;DR:** - AI agents are already making real payments - Mastercard just proved it across multiple markets. - This changes how payments need to work - systems must support both AI-driven flows and human checkout. - Missing or misconfigured payment methods can break automated workflows instantly. - We shipped [v1.93.0](https://docs.dodopayments.com/changelog/v1.93.0#1-checkout-redesign) to help with both - better control over payment methods and a smoother checkout experience. - Also, we're introducing a monthly builder riddle, solve each one and get them all right by the end of the month to earn a reward. **Hello everyone,** I was reading about Mastercard's latest rollout last week, and it didn't feel like a typical "future of AI" announcement. They've already processed real payments through AI agents - not demos or internal tests, but actual transactions like groceries, digital goods, and even rides, completed end-to-end across Latin America and Hong Kong. No one clicking "Pay." No manual confirmation. Just agents executing transactions using tokenized credentials and passkeys. And that's when I thought, this isn't something we're building toward anymore. It's already happening. ## What Changed Overnight For the past year, most of us have been talking about AI agents handling payments as something that's just around the corner - a logical next step as agents get better at reasoning and executing tasks. What Mastercard showed this week is that the step isn't theoretical anymore. It's already happened. Their Agent Pay setup is live across multiple markets, with banks and issuers supporting real transactions. AI agents can now discover, select, and execute payments across different regions without relying on a traditional checkout experience. And that shift is more important than it initially looks. Because once payments move from a UI-driven flow to an API-driven one, the underlying structure of your system starts to matter far more than the interface itself. An AI agent isn't navigating your checkout the way a user would. It's querying your system, understanding what options are available, and selecting from what you've exposed. Which means the quality of your payment layer is no longer just about what the user sees, it's about what your system allows. ## Where Things Break This is where things get interesting. When an AI agent interacts with your system, it doesn't have the context a human does. If your setup exposes a payment method that doesn't work well in a specific region, the agent won't question it, it will simply try it, and the transaction might fail. If a required payment option is missing for a market, the workflow doesn't adapt - it just stops. In other words, the system behaves exactly as you've defined it, with no room for intuition or correction. And that makes the structure of your payment layer far more critical than it used to be. At the same time, while all of this is happening on the backend, the frontend reality hasn't changed. Humans are still showing up at checkout, and they're still making decisions in the same way they always have. They hesitate when something feels off, they drop off when the experience is slow or confusing, and they judge your product based on how that final step feels. So now you're operating in two parallel realities. On one side, AI agents need clean, reliable, programmable payment rails that don't break under edge cases. On the other hand, human users still need a fast, intuitive, and trustworthy checkout experience. Most systems today weren't designed to handle both at the same time and that gap is where things start to break. ## How to Think About This If you're building today, your payment layer needs to do two things well. First, it needs to be controllable. You should be able to decide which payment methods are available globally - not at runtime, not per session, but at a system level. Second, your checkout experience still matters. Even if AI handles part of the flow, your human users are still making decisions at the end. Slow loads, broken states, or confusing UI still cost you conversions. Third, trust can't just be visual anymore. It needs to exist in the system itself - through authentication, tokenization, and predictable behavior. And finally, you need clarity. Not just total revenue, but what's actually happening underneath, after refunds, disputes, and fees. Because without that, you're optimizing blindly. ## What We Shipped (And Why It Matters) This is exactly what we focused on in our latest release. We realized that as systems become more automated, control and clarity become more important than ever. So with [v1.93.0](https://docs.dodopayments.com/changelog/v1.93.0#1-checkout-redesign), we introduced [business-level payment method control](https://docs.dodopayments.com/changelog/v1.93.0#3-business-level-payment-method-disabling). Instead of managing payment options per session or per flow, you now control everything from one place. You decide which methods are available across your entire system instantly, without code changes. For AI-driven workflows, this means your system only exposes methods that actually work. For human users, it means a cleaner, more relevant checkout. Alongside that, we rebuilt [checkout](https://docs.dodopayments.com/features/checkout) itself. No flickers. No awkward loading states. A smoother flow, with clear navigation and exit paths, because even if AI is driving transactions in the background, your users are still experiencing that final step. And we added something simple, but important. [Business-level control over payment methods.](https://docs.dodopayments.com/changelog/v1.93.0#8-donotbill-proration-mode) You can now disable specific payment methods across your entire business from one place, without configuring it per product or session. This matters more than it sounds, because when an AI agent (or even a user) hits your checkout, it will only work with what you expose. If the wrong method shows up, or the right one is missing, the flow breaks instantly. So instead of reacting per transaction, you define it once at the system level, and everything stays consistent. Alongside this, we also expanded support for **Pix and WeChat Pay**. Pix dominates payments in Brazil, and if you're not supporting it, you're effectively invisible there. And in China, people don't adapt to your checkout - your checkout has to adapt to them. Because reaching global users isn't just about being available. It's about being payable. ## One Last Thought What Mastercard showed this week isn't just a technical milestone. It's a shift in how commerce happens. AI agents are now part of the transaction flow. But they're not replacing users, they're running alongside them. And the systems that win won't be the ones that only work for one side. They'll be the ones that handle both seamlessly. And before we wrap - we're trying something new this month. Each issue will have a small riddle for builders. Solve them as we go. _"I'm the reason a user stays, but you'll never see me in your code. I'm felt in a second, lost in a glitch, and once I'm gone, the sale is gone too."_ Got the answer? Hold onto it till the last edition of April. If you get all of them right, there's a reward waiting. Also, join our loving [Discord community](https://discord.gg/dodo-payments-1305511580854779984)! Best, Rishabh Goel Co-Founder, **Dodo Payments** --- - [More Newsletter articles](https://dodopayments.com/blogs/category/newsletter) - [All articles](https://dodopayments.com/blogs)