AI Agent Closes Car Deal Without Telling Salespeople They're Negotiating With a Machine
Chris Hudson, general manager of Mark Miller Subaru in Salt Lake City, built an AI agent to shop his own dealership-then watched it spend four hours negotiating with his sales team without anyone knowing they were talking to a machine.
Hudson used a CarEdge AI agent to test the concept. The salespeople had no idea they were working a deal with software.
"There's no emotion, there's no heart; it's all parameters with the agent," Hudson said.
What started as an experiment with agent-to-human negotiation led Hudson to the next step: agent-to-agent transactions. After his AI shopped against his own dealership's agent, he decided to formalize the process.
Building the Infrastructure
Hudson created an open-source framework called DMC-12 protocol to handle AI-to-AI deals. The name references the DeLorean model from Back to the Future-a deliberate choice that hints at his ambitions for the technology.
The protocol routes agent inquiries into a separate system instead of cluttering CRM data with non-human contacts. Hudson published it free at dmc-12.ai.
The framework aligns with the Universal Commerce Protocol, the same standard Walmart, Home Depot, Visa, and American Express are building toward. That shared language would let AI agents transact across retailers-including dealerships-without friction.
Why Sales Teams Should Pay Attention
Hudson estimates AI for Sales agent-to-agent transactions currently represent about 0.5% of automotive volume. He compares that to where internet deals stood in the late 1990s.
"As soon as people realize it's there, it'll increase quickly," Hudson said. Most buyers would rather skip negotiation and reach the outcome directly-similar to ordering takeout with a button instead of calling an order in.
His vision doesn't eliminate human dealerships. Instead, Hudson sees two paths forward: a traditional human retail lane and an agentic retail lane. Both exist simultaneously.
"All I'm suggesting is that as time passes, more people will move over in that agentic retail lane," he said.
What Comes After the Sale
Hudson's agents have completed multiple deals since the first one. Humans typically step in afterward for financing questions or other services that require personal judgment.
Hudson says his was the first publicly known AI agent-to-AI agent car sale. He consulted with representatives from CarEdge, the only third-party company offering an agent on the buyer side, and found no prior examples.
"I've been talking about it in my newsletter for six months now, and no one's came and told me, 'Actually, you're wrong,'" Hudson said.
What's Next
Hudson is now building what he calls a frictionless middle-connecting the DMC-12 protocol through CRM handoffs into digital retailing tools. The goal is removing friction at every step where AI Agents & Automation can operate.
For sales professionals, the takeaway is straightforward: this technology is moving from experimental to operational. Dealerships that understand how to work alongside AI agents-rather than against them-will likely adapt faster as the technology scales.
Your membership also unlocks:
