A common question about agentic commerce is whether it will fragment into competing ecosystems.

Will there be an AI-led version? A platform-led version? A payments-network-led version? Or several approaches that coexist?

In the early stages, the answer is almost certainly yes.

Different players are approaching the problem from where they sit. AI companies focus on agent behaviour and coordination. Commerce platforms focus on merchants, catalogues, checkout, and distribution. Payments providers focus on authorisation, settlement, and liability.

Fragmentation at this stage isn’t a failure. It’s how new execution models emerge.

What’s important is where fragmentation persists — and where it doesn’t.

Execution layers can diverge without breaking the system:
• How agents are invoked
• How interactions are structured
• How checkout is rendered
• How tools are called

Different protocols and frameworks will compete and evolve here.

But fragmentation becomes much harder to sustain when it comes to trust, authority, and consent.

As soon as software is authorised to act — to spend money, share data, or commit an organisation — the same questions surface everywhere:
• Who authorised this?
• What limits applied?
• What information was disclosed?
• Can that authority be revoked?
• Can the outcome be independently verified later?

Those questions don’t belong to any one ecosystem. They belong to users, merchants, payment networks, and regulators.

Execution can be proprietary.

Trust can’t be.

Understanding where fragmentation is acceptable helps narrow the real problem. Once software is authorised to act, the challenge isn’t coordination between systems — it’s how responsibility is delegated, constrained, and evidenced when things go wrong.