A Listed Car Retailer Just Put AI On Its Earnings Call. The Biggest Lever Wasn't On It.

# A Listed Car Retailer Just Put AI On Its Earnings Call. The Biggest Lever Wasn't On It.

An 83% profit jump, and AI got the credit.

In results published in June for its year to March, a listed UK car retailer reported pre-tax profit up 83% to £7.5 million, revenue past £1.27 billion, and a record 64,600 retail units. Management put the story squarely on data and AI: algorithm-based vehicle allocation, data-led pricing, and an agentic system that re-engaged customers sitting on closed quotes — credited with 877 incremental sales. Customer acquisition cost fell to £163 a unit, from £177.

Read the whole release and you notice something. Every AI win they published lives on the same side of the store: sourcing, pricing, marketing, reviving a lead that already raised its hand. The buy side and the sell side.

The side that never made the earnings call is the phone.

The efficiency lands wherever the work was already digital

This is not a knock on them. Those are real, hard-won gains, and reviving a dead quote with an agent is exactly the kind of unglamorous automation that compounds. It is worth pointing out because it is the pattern in almost every "AI made us more efficient" story published this year. The efficiency shows up wherever the work was already digital. Where it does not show up is the one channel that stubbornly stays human and stubbornly stays analog: an inbound call to book service.

That is a strange blind spot, because the phone is where the highest-intent customers are — and the data on that is not subtle.

Phone callers convert. Then they hit voicemail.

Foureyes' dealership data found phone leads set appointments at about 74%, versus roughly 40% for internet leads, and close at about 2.3 times the rate. A caller is someone who wants to give you money badly enough to interrupt their day and dial. And in most stores, at the exact moment that person rings to book a repair, the outcome is a hold, a voicemail, or a callback that never comes — because the advisors who could book them are already elbow-deep in the customers standing at the counter.

So the same operator who will spend six months and real budget wringing efficiency out of a pricing model will let a fully-formed, ready-to-book service customer roll to voicemail at 8:52am. Not because they do not care. Because the phone has never had an owner, and you cannot put "we missed fewer calls" on a slide unless someone was counting.

The math that reframes it

In a franchise service drive the average repair order runs around $400, and the missed-call problem is not a rounding error: the recoverable service revenue a single rooftop leaves on the table from unanswered and after-hours demand runs to roughly $450,000 a year. That is not a marketing number. It is the gap between the appointments a store books and the ones it could have booked if every call that came in got answered and turned into a written-back appointment instead of a lost one.

This is where voice AI earns its place — and where it has to be honest about what it is and is not. Answering the service phone is not the flashy agentic story. It is the boring one: pick up on the first ring, at 8:52am or 8:52pm, in the caller's language, book against the real schedule, and write it back to the dealer management system so an advisor is not retyping it. Done right, a booking that takes an advisor twenty minutes between customers takes the machine about ninety seconds — and, the part that matters, the caller never knew they were in a queue, because there was not one.

The two-pile test for your AI roadmap

If you run a store or a group, here is the test to apply to your own AI roadmap this year. Take every AI initiative you are funding and sort it into two piles:

• The ones that make already-digital work cheaper.
• The ones that stop a ready-to-buy customer from falling through a crack.

Almost everyone is over-invested in the first pile. The first pile is easier to measure and easier to demo. The second pile is where the retention actually leaks.

So before you approve another pricing model, pull one number. How many inbound calls did your service department miss last month, and what is a booked repair order worth to you? Multiply them. That is the line that never makes the earnings call — and for most stores it is bigger than the ones that do.