

Most AI "case studies" in this industry are a logo, a quote from a VP of Innovation, and a stat that conveniently can't be checked. This week gave us one that's different enough to slow down and read carefully — which is exactly what we did, twice.
Cars24 turned over roughly 1 million minutes a month of customer calls and chats — across used-car buying and selling — to OpenAI-powered voice and chat agents. The results, as reported: 50% faster support resolution, 80% less staff time spent on appraisal busywork, and 12% of sellers who'd already ghosted the funnel pulled back in and re-engaged.
That last number is the one that should make an operator sit up. Every store has a graveyard of leads that went quiet — trade-in inquiries that never got a callback, appraisal requests that stalled because nobody had the bandwidth to chase them. Cars24's bot didn't just answer new calls faster; it went back and worked the dead pile, and a chunk of it came back to life. That's not a chatbot demo. That's found revenue.
Why this matters for your store: the credible part of this story isn't "AI can talk to customers," which we've heard for two years now. It's that the volume was real (a million minutes doesn't happen in a sandbox), the staff-time savings were measured against actual appraisal workflows, and the recovery rate is a number a skeptical GM could ask to see broken down by month. Most vendor pitches can't survive that question. This one, so far, has.
The caution flag: it's still a vendor-adjacent case study, and "50% faster resolution" doesn't tell you what resolution quality looked like, or whether customers who got bounced to a human were counted as failures or successes. Ask your own vendor for the denominator before you believe the numerator.
If voice AI can reliably resurrect 1 in 8 dead leads, the next question isn't whether to deploy it — it's whether your CRM is clean enough to give it something worth resurrecting. Garbage lead data in, garbage recovery rate out. Before you sign anything, ask what happens to the 88% who didn't come back.

This week's evidence says the AI isn't the variable — your setup is. One dealership tripled its group's average contact-rate gain on identical software, while mystery shoppers found customers juggling four different chatbot "names" before a human ever picks up the phone.

AI is now answering dealership leads at 10 p.m. and booking appointments before sunrise — but a mystery shopper found customers getting bounced between four different "people" in their first 72 hours. The chatbots are getting faster. The handoff to an actual human still isn't built, and that gap is where deals quietly die.