There is a number from C-4 Analytics' Quarterly Consumer Intent Survey that I keep coming back to: only fourteen percent of car buyers trust AI as much as a dealership salesperson.
That is a trust gap. And it is a real problem for dealers who are deploying AI tools across their BDC, their inventory management, and their digital marketing.
But here is the number that concerns me more: the percentage of dealers who know what their AI is actually doing.
I do not have that number. Nobody does. Because most dealers do not have visibility into their AI tools at a level that would let them answer the question.
The Black Box Problem
When a dealer deploys an AI voice agent to handle inbound calls, they typically see two metrics: call answer rate and appointment booking rate. If both go up, the tool is working. If they go down, the tool is not working.
What they do not see is what the AI is saying. They do not see the conversations. They do not see the cases where the AI gave a customer incorrect information about a vehicle's availability, or quoted a price that was out of date, or failed to recognize that a caller was a returning customer with a service history.
Those failures are invisible unless you are actively monitoring them. And most dealers are not actively monitoring them, because monitoring AI conversations at scale requires tooling that most dealers do not have.
The Confidence-Competence Gap
This connects to something I wrote about in the context of knowledge engines: the difference between confidence and conviction. A system can be highly confident and still be wrong. And a system that projects confidence it has not earned is more dangerous than a system that admits uncertainty.
AI voice agents are, by design, confident. They do not say "I am not sure." They give answers. And in a high-stakes transaction like a car purchase: where a customer might be making a $50,000 decision based on information they received from an AI: a confident wrong answer is worse than no answer at all.
The dealers who are winning with AI are not the ones who deployed it and moved on. They are the ones who built monitoring into the deployment. They are listening to AI conversations. They are tracking cases where the AI's answer did not match reality. They are using those cases to retrain the model and tighten the guardrails.
That is not a set-it-and-forget-it deployment. That is an ongoing product management process.
What Good AI Governance Looks Like at a Dealership
Based on what I saw in the C-4 Analytics market intelligence work, the dealers who are managing AI well are doing a few things consistently.
They have defined the boundaries of what the AI is allowed to say. Not just what it can say, but what it cannot say: specifically around pricing, availability, and financing terms. Those are the high-stakes areas where a wrong answer creates real liability.
They are reviewing AI conversations on a sample basis every week. Not every conversation: that is not scalable: but a random sample large enough to catch systematic errors.
They have built a handoff protocol. When the AI encounters a question it cannot answer confidently, it hands off to a human. The handoff is seamless from the customer's perspective, but it is a critical safety valve.
And they are tracking the handoff rate as a product metric. A high handoff rate means the AI's guardrails are too tight. A low handoff rate might mean the AI is answering questions it should not be answering confidently.
The Trust Gap Is a Product Gap
The fourteen percent trust number is not a consumer education problem. It is a product problem. Consumers do not trust AI because AI has given them reasons not to trust it: confident wrong answers, impersonal interactions, the uncanny valley of a system that sounds human but does not quite understand context.
Closing that gap requires building AI tools that are honest about their limitations. That ask for help when they are uncertain. That surface their reasoning, not just their conclusions. That earn trust through accuracy, not just through confidence.
That is harder to build. But it is the only path to a world where AI in automotive retail actually serves customers well: rather than just making dealers feel like they are keeping up with the technology curve.
