𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭 𝐓𝐫𝐮𝐬𝐭 𝐋𝐚𝐲𝐞𝐫 𝐖𝐨𝐧’𝐭 𝐁𝐞 𝐚 𝐂𝐨𝐦𝐦𝐨𝐧𝐬. 𝐈𝐭 𝐖𝐢𝐥𝐥 𝐁𝐞 𝐚 𝐏𝐫𝐨𝐝𝐮𝐜𝐭.

Last week I wrote that nobody would pay to build the trust infrastructure AI agents will need to transact.

I was wrong.

Several organisations have already announced versions of it.

The problem is simple.

As AI agents start booking, buying and paying for us, businesses need answers to three questions:

Who is this agent?
Who is it acting for?
What is it allowed to do?

The infrastructure answering those questions is already appearing.

April 30th, Experian, the FTSE 100 credit bureau, announced Agent Trust, linking a person to the agent acting for them and evaluating its behavior over time.

Eight days later, three Chinese ministries proposed exploring a national agent registration platform with digital identities and declarations of what agents can do.

Ant Group, the financial arm of Alibaba, followed with technology for safer agent-initiated payments.

JD.com, the Chinese e-commerce giant, announced its own approach to verifying the user, the agent and its operating environment when payment occurs.

These are different systems. But they solve the same problem:

Before an AI agent can transact, someone needs to decide whether it can be trusted.

That changes my original argument.

I was right that a public good with a private cost is difficult to build.

I was wrong to conclude that the infrastructure therefore stays unbuilt.

It does not arrive as a commons.

It arrives as a product, a platform or regulation.

The incentives are clear.

Payment providers need to prevent fraud. Platforms need to decide which agents can transact. Identity companies can sell trust. Governments may require registration.

China shows where this could lead.

As of 30 April of this year, 868 generative AI services had completed regulatory filing there.

Those are AI services, not registrations in the proposed agent platform. But they show how registration can become part of the permission to operate.

So the executive question is not:

“Which agent standard will eventually win?”

It is:

“Which systems are our partners choosing for us today?”

Ask your payment provider how it will identify AI agents.

Ask your commerce platform which standards it is adopting.

If China matters to your business, ask what registration agents may eventually require.

These sound like technical decisions.

They are not.

If AI agents become customers and buyers, whoever controls the trust layer gets a say in who can transact.

The question is no longer whether that layer will exist.

It is who will control it and how much of that decision is still yours.

That is the next post.

#AgenticCommerce #ArtificialIntelligence #DigitalIdentity #RetailStrategy #ChinaBusiness

Contact

bruno.gentil@sherpaconsultingasia.com

© 2026. All rights reserved.