I want to start with something that has nothing to do with cars.
A few years ago, one of my personal accounts on a very large social platform was flagged and restricted by an automated system. The reason, which took considerable effort to establish, was that a batch of listings had been misclassified as an adult product category. They were decorative stones. Elongated, granted. Not that elongated.
The account happened to be the one that administered a business advertising account, so this was not a small inconvenience. So I did what you do. I contacted support.
What followed took months. I filled in forms. I spoke to people on video calls. I explained the situation to a first-line agent, who escalated it. I explained it again, from the beginning, to the person they escalated it to, who escalated it further. I explained it again. And again. At every level I started from zero, because nothing I had said before arrived with me.
Every single one of those interactions was with a human being. Not one of them was capable of helping. The eventual resolution, after all of it, was a suggestion that I create a new account.
The thing that experience taught me
It would be easy to read that as a story about bad customer service, and it is, but that is not the interesting part.
The interesting part is that it was inhuman while being staffed entirely by humans.
The people I spoke to were not unkind. Several of them were clearly trying. They were also completely powerless - buried in organisational debt, working in a process where no individual had the authority or the visibility to fix the thing, and where the escalation path was a way of moving a problem rather than solving it. Being handled by a person turned out to be no guarantee whatsoever of being treated like one.
And here is the comparison that matters. In the last year I have had several dealings with AI support agents at large companies, knowing full well what I was talking to. Account issues, technical settings, things switched off that should have been on. Ordinary trivia.
They were resolved. Quickly, on the first attempt, with full memory of what I had already said.
I did not want a relationship with those companies. I wanted to get from A to B. The machine got me there and the humans did not, and any theory of what customers want that cannot accommodate that fact is not a theory about customers. It is a theory about how we would prefer the world to be.
What people actually want
The objection to AI in customer contact is almost always framed as a matter of respect: customers deserve a human being.
Sometimes that is exactly right, and I will come to when. But it conflates two very different kinds of interaction, and the conflation is where the argument goes wrong.
| INTERACTION | WHAT THE CUSTOMER WANTS | WHAT MATTERS MOST |
|---|---|---|
| Getting a fact established: is it available, what does it come with, what does your warranty say, when can I come in | To get from A to B, now | Speed, accuracy, memory of what was already said |
| The decision itself: valuing the trade-in, sitting in the car, negotiating, committing | To be understood and to trust the person opposite | Presence, judgement, warmth |
Most of what arrives in a dealership’s inbox is the first kind. Somebody wants to know whether a car is still there, what it comes with, what the warranty covers, and whether they can see it on Saturday. None of that requires warmth. It requires a correct answer, promptly.
The second kind is where the sale actually happens, and it is where a person is irreplaceable. Simon-Kucher’s 2025 global study found 82% of buyers still prefer to finalise their purchase at the dealership.
The mistake the industry makes is spending its scarce human capacity on the first kind, and then having none left for the second.
The honest comparison
The argument against AI in dealership enquiries usually compares it to a good salesperson giving a thoughtful reply.
That is not the comparison. The comparison is with what actually happens.
58% of consumer enquiries arrive outside opening hours, peaking between 21:00 and 23:00. When AutoUncle asks the people who sent them, 26% report they still had no reply two days later.
So for a substantial share of buyers, the real choice is not between a machine and a person. It is between a machine at 22:15 and nothing at all until Tuesday, by which point they have written to three other dealerships and one of them has answered.
Nobody is choosing warmth over efficiency in that scenario. They are choosing between an answer and silence, and silence is not the more human option merely because no machine was involved in producing it.
The thing your own staff cannot do
There is a second argument here that is less about philosophy and more about a practical limitation nobody likes to state plainly.
Ask yourself how many of your salespeople could accurately describe your current warranty terms. How many know what changed on your website last month. How many could tell a customer, without checking, what your policy is on unaccompanied test drives, or which of your locations closes early before a bank holiday.
In a larger group this gets worse, not better. Decisions are made centrally and communicated once, in a document, to people who were with customers when it arrived. Everybody means well. Almost nobody has read it.
This is not a criticism of salespeople. It is a straightforward observation about human memory and the volume of things a dealership decides in a year. A well-built system does not have that problem - not because it is cleverer, but because it is reading the current version every time rather than remembering an older one.
Which produces a genuinely uncomfortable conclusion. On settled questions of fact about your own business - what the policy says today, not what it said in March - a properly grounded system is working from the current version, while everyone else in the building is working from whenever they last read it. And it is doing so at 22:15 on a Saturday, when the person who would otherwise answer is asleep.
The part that would make it inhuman
None of this is an argument for putting AI everywhere, and the failure mode is real enough that it deserves naming.
An AI system deployed to deflect customers rather than to serve them is exactly the thing I described at the start of this article, rebuilt more efficiently. If it exists to keep people away from a resolution, it does not matter how well it writes. It is the escalation carousel with better grammar.
The tests for the difference are not subtle:
Does it resolve, or does it hold? A system that answers questions and books appointments is serving. A system whose main function is to absorb contact until the customer gives up is not.
Does it know when to stop? Anything involving money, liability or commitment - price, negotiation, discounts, putting a figure on a trade-in, financing terms, interpreting warranty or insurance conditions, diagnosing a fault, guaranteeing availability - should hand over to a person, quickly and without being asked twice.
Is it honest about what it is? Since 2 August 2026, EU law requires this: Article 50 of the AI Act mandates clear disclosure at the start of the first interaction, with fines up to EUR 15 million or 3% of global turnover. But it was the right call before the law arrived. A system maintaining a disguise cannot be straightforward about its own limits. One that has nothing to hide can say plainly: the car is available, Wednesday at 13:00 looks good and I will get it confirmed, and a colleague will talk to you about the price. There is no character to break.
Where this ends up
The version of this that works is not a dealership with fewer people in it. It is a dealership where the people are doing the part that needed a person.
Somebody stands there with the keys on Wednesday afternoon. They already know the budget, the trade-in, and what the customer has been looking at, because that conversation started on Saturday night while everyone was asleep. They spend the visit on the car and on the person, rather than on the eleven factual questions that could have been settled days ago.
That is a more human dealership than the one where a tired salesperson triages forty ambiguous messages on a Monday morning and answers eight of them.
Being staffed by people has never been the same thing as treating people well. It is what those people are freed up to do that decides it.
Frequently asked questions
Will customers object to dealing with an AI? Some will, and they should be able to reach a person immediately. But the honest comparison is not AI versus a thoughtful human reply - it is AI versus no reply until Tuesday, which is the actual outcome for a substantial share of enquiries. Most people evaluating a purchase in the evening want an answer more than they want a particular kind of correspondent.
Is AI customer service worse than human customer service? It depends entirely on the alternative. Large organisations routinely deliver human-staffed support that resolves nothing, because the people in it lack the authority and context to help. A grounded AI system that can actually answer will often outperform it. For decisions requiring judgement, trust or negotiation, a person is not replaceable.
How do I keep AI from making my dealership feel impersonal? Use it where facts are being established and keep it out of where relationships are built. Make handover to a person fast and easy. Give it hard limits on price, negotiation, valuations and any other commitment carrying money or liability. Disclose what it is.
Does the EU require disclosing AI to customers? Yes. Article 50 of the EU AI Act has applied since 2 August 2026 and requires that a person interacting with an AI system is told so clearly, at the start of the first interaction, with fines of up to EUR 15 million or 3% of global turnover.
Does AI replace salespeople in a dealership? It replaces a part of the work, not the role: reading and triaging ambiguous enquiries, chasing missing details, answering routine factual questions out of hours. The test drive, the price conversation and the handover are where the sale is made and where buyers want a person.
Sources
- Share of enquiries arriving outside opening hours (58%, peak 21:00-23:00): AutoUncle data, 2026.
- Share of consumers reporting no reply two days after enquiring (26%): AutoUncle lead survey, 2026.
- Preference for finalising the purchase at the dealership (82%): Simon-Kucher, Automotive Shoppers Take the Wheel, global consumer study, 2025.
- EU AI Act Article 50 transparency obligations, applicable from 2 August 2026: European Commission.
