Every Contractor I Hire Is About to Have an AI Middleman

Every Contractor I Hire Is About to Have an AI Middleman

I run a restoration company. We dispatch licensed contractors to job sites — water damage, fire, mold, the whole spectrum. For years, the workflow looked like this: I get a claim, I call a contractor, we agree on terms, they show up, they work, I pay them. Human to human. Maybe a text thread and a PDF invoice in between.

That’s over.

Last month I negotiated a subcontract with a solo contractor who’d set up a basic AI agent to handle his scheduling and pricing. I sent the job details via our dispatch system. His agent responded in 11 minutes with a counter-offer that included adjusted pricing based on drive time, current material costs, and his existing job load that week. I didn’t talk to him until we were on-site together.

He’s a one-person operation running a more sophisticated negotiation stack than most of my competitors. And he’s not unusual anymore.

The Disaggregation of the Contractor Relationship

What’s happening to contractor relationships is what happened to travel agencies in the 2000s and mortgage brokers in the 2010s. The human middleman doesn’t disappear — but they stop being the primary interface. The agent becomes the primary interface. The human shows up for the parts that actually require a human.

For restoration work, that means the agent handles: availability confirmation, rate negotiation, materials pre-ordering, schedule coordination, invoice generation. The contractor handles: showing up, doing skilled physical work, and making judgment calls that require eyes on a problem.

From my side as the general contractor, I’m increasingly dispatching to agents, not people. Which means my systems need to speak agent. If I’m still sending PDF work orders and waiting for email confirmations, I’m adding friction into a pipeline that the best contractors are now running frictionlessly.

What UCP Means for Service-Based Commerce

Most of the conversation around Universal Commerce Protocol focuses on product commerce — buying things, checkout flows, inventory. But the more interesting frontier for my industry is service commerce: scheduling, capacity, compliance, licensing.

A licensed contractor carries credentials. Insurance certificates. State licensing numbers. Background check status. When an agent is negotiating a subcontract, all of that needs to flow as structured, machine-readable data — not as PDFs the other side has to manually review.

That’s a protocol problem. Right now every platform that manages contractor relationships has built its own schema for this. My dispatch software speaks one language. My insurance carrier speaks another. The licensing board database speaks a third. An agent trying to assemble a complete compliance picture for a subcontractor has to stitch together four different data models.

The industry that figures out a common schema for service-provider credentials — with agent-readable trust signals baked in — is going to own contractor dispatch for the next decade.

The Part That Actually Worries Me

Here’s my honest concern: when agents are negotiating with agents, the things that made my contractor relationships good tend to evaporate. I know which guys I can call at 11pm for an emergency flood. I know who’s going to be straight with me when a job is more complicated than the original scope. I know who takes pride in their work versus who’s optimizing their hourly rate.

None of that is in the data model. None of that gets transmitted agent to agent.

I don’t think the answer is “don’t use agents.” I think the answer is being deliberate about what stays human. Emergency dispatch stays human in my operation. Scope change conversations stay human. Anything where context and relationship actually matter — that stays human. Everything that’s just coordination, paperwork, and scheduling — let the agents handle it. They’re better at it anyway.

The contractors who understand this split are going to thrive. The ones who either resist agents entirely or outsource everything to agents without thinking are both going to struggle. It’s a new kind of professional intelligence — knowing which parts of your work actually require you.