The protocol debate peaked my interest. The question underneath it is more interesting.
When content providers start requiring that agents pay for API calls, web content, and compute on your behalf, where does that cost land?
Right now, most AI costs are visible. You pick a model, you see token pricing, you set a budget. But agentic payments add a new layer. Your agent calls a service, that service charges for access, and the cost accumulates across dozens of micro transactions. Are we moving towards agent owners needing to approve agent budgets?
Two things need real solutions before this works at scale. A standard way to distinguish human traffic from agent traffic. And what it actually costs to let an agent use a resource. That second one matters most.
Operational costs are shifting from human labor to digital transactions. That shift happens because digital is cheaper. But cheaper is not free. And margins on digital services are compressing, which means providers will need to recover costs somewhere. The question is whether that somewhere is the agent operator, the end user, or some new pricing layer that doesn't exist yet.
We spent years figuring out SaaS pricing. Per seat, per usage, per feature. Agent commerce breaks most of those models because the "user" making the request isn't a person. It's a process that might make 10,000 requests before a human sees any of it.
How do you think agent resource costs will eventually reach the end user?
Written by Duane Grey
AI Strategy & Implementation
Independent AI consultant helping companies cut through hype and deploy systems that produce real results.