Why I'm Building in Agentic Commerce Right Now

I have worked in payments for a long time. I have seen many "next big things" come and go. So I do not say this lightly: agentic commerce feels different. I think it is worth building on now, not later.

This post is not a product pitch. It is my honest view on why I am spending my time here, and where I think the real problem is.

What changed for me

For years, the story was about better checkout buttons and faster pages. Small steps. Then AI agents became able to do a full task: plan, use tools, and act.

The first time I watched an agent complete a real purchase — find the item, fill the cart, and pay — something clicked. The buyer was not a human clicking. It was software acting for a human. That is a new kind of customer. And a new customer changes everything about how a store must work.

The hard part is not the AI

Here is the thing many people miss. The AI is not the hard part anymore. The models are already good enough.

The hard part is money and trust. Three questions keep coming back:

• How do we prove a human really approved a purchase?
• How does a store know an agent is real and not a fake?
• Who pays when an agent buys the wrong thing?

I have spent my career close to these exact problems — tokens, authorization, fraud, disputes, and strict compliance. So when I look at agentic commerce, I do not see a magic AI problem. I see a payments and trust problem wearing new clothes. That is a problem I know.

Why now, and not in two years

Some people say it is too early. I understand why. The standards are still moving. Some announcements are just plans, not shipped features.

But that is exactly why now is the right time. The rules are being written this year. Card networks, protocols, and standards bodies are all deciding how agents will pay and prove themselves. If you build while the ground is still soft, you can help shape it. If you wait until it is set, you only get to follow.

I would rather be early and useful than late and safe.

What I am building

I am building Sarvex, an operating system for agentic commerce. The simple goal is to give merchants and builders the safe layer they need so an agent can shop and pay without chaos: clean product data agents can read, safe checkout, agent identity, and control over spending.

I am not going to pretend it is finished. It is not. I am building in the open, one piece at a time, and I will write about the real lessons here — including the mistakes.

What this blog will be

This blog has three kinds of posts:

• Reference posts that explain the concepts in plain language.
• Signal posts that map what is happening in the industry.
• Field notes like this one, where I share what I am learning as I build.

I write in simple, direct language on purpose. English is not my first language, and I think clear writing helps everyone, including me.

If you are building in this space, or just trying to understand it, follow along. I will keep it honest, and I will keep it useful.

More soon.