The Moment a Customer Conversation Becomes Just a Lead

A conversation that works, followed by a system that doesn't

A customer spends several minutes talking to an AI assistant on a dealership website. They explain exactly what they are looking for. They mention the vehicle they want to trade in and what they still owe on it. They ask whether they can keep their monthly payment under a certain number and whether financing is something they can sort out when they come in.

It is a genuinely useful conversation. The customer leaves feeling like the dealership actually knows something about them.

Then that interaction enters the dealership's systems, and most of what just happened gets thrown away.

What reaches the CRM looks something like this: name, phone number, vehicle of interest. The trade-in details, the payment range, the appointment preference, the financing question -- all of it is gone. Not because the AI failed to capture it. It captured everything. The problem is the systems that carried the interaction were not built to hold that kind of information. So they did not.

Infrastructure built for a different era

For about two decades, most automotive leads arrived through simple web forms. A visitor typed in their name, email, phone number, and the vehicle they were interested in, and that data moved into a CRM through a format called ADF, the Auto-Lead Data Format. ADF was well designed for what it was asked to do. It gave the ecosystem a shared language, and the industry adopted it widely.

The problem is that the customer interaction has changed fundamentally, and the infrastructure has not kept up. A form submission and a 20-message AI conversation are completely different things. One produces four fields. The other produces a detailed picture of who the buyer is, what they want, what they can afford, and when they are ready to act.

Routing both through the same infrastructure and expecting the same result does not make sense anymore.

The context is there, but the infrastructure loses it

I want to be clear about something. This is not a criticism of ADF, and it is not a criticism of the CRM platforms that rely on it. Those systems do exactly what they were designed to do. The issue is that nobody has built the layer that sits between a modern AI conversation and those existing systems -- something capable of carrying richer interaction data forward without stripping it down to a handful of fields.

When that layer is missing, every step downstream starts with a degraded version of the truth. The salesperson who picks up the phone knows the customer wants a RAV4. They do not know about the trade-in payoff, the payment target, the financing question, or the fact that Saturday at 10am was already discussed. So they treat it like a cold lead, the customer has to repeat themselves, and the value of the AI conversation evaporates.

In the AI Operator Framework, this is a Govern phase failure. The system was never governed to preserve context across handoffs. There are no quality gates between the AI output and the CRM input. No rules about what data must survive translation. No measurement of what gets lost. Without governance, every integration becomes a silent data leak.

Building the missing layer

I have spent the last year building the Automotive Commerce Protocol -- ACP -- which is an attempt to define what that missing layer could look like. It is an open, JSON-based specification designed to carry conversation-level interaction data between systems while staying backward compatible with the ADF pipelines already in place.

ACP wraps every interaction in a structured envelope that preserves intent signals, deal stage, trade-in context, payment targets, appointment data, and the full conversation transcript. When the lead moves from AI assistant to CRM to salesperson to F&I, the envelope travels with it. Every downstream system picks up where the conversation left off instead of starting over.

The goal is not to replace anything. It is to add the layer that is currently missing.

The question that matters

AI is going to keep getting better at having useful conversations with customers. The models will improve, adoption will grow, and the conversations will carry more context than they do today.

But if the infrastructure those conversations flow into keeps treating them like form submissions, the industry will keep discarding the most valuable part of what AI actually produces. Better AI on top of the same infrastructure does not solve the problem. It just makes the gap more expensive.

When your AI has a great conversation with a customer, where does that conversation actually go?