

What's new: Cars24, the used-car marketplace, published results from a year-plus rollout of OpenAI-powered voice and chat agents — the system now handles over 1 million customer conversation minutes a month, and it's not just answering questions.
Why it matters: Most "AI customer service" claims from dealers and platforms are vibes and demo videos. This one has numbers attached, and they're the kind operators actually care about — resolution speed, staff hours, and recovered deals that would otherwise walk.
The catch: This is Cars24's own case study, distributed via OpenAI — not an independent audit. There's no word on cost per conversation, error rates, or what happens when the agent hits something it can't handle. Treat the topline numbers as a strong signal, not gospel, until a third party or a competitor replicates them at similar scale.
Bottom line: If you're still routing appraisal scheduling and lead follow-up through a human phone tree, this is the kind of proof point that should move that project up your list — not because AI is magic, but because recovering abandoned sellers is a number your GM already tracks.

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.

Flai closed a $27M Series A after its AI agents started booking 50,000 dealership appointments a month and posting a 20x revenue jump, with backers including Toyota Ventures and two dealer groups putting real money behind it. But the same week, a veteran consultant warned that unsupervised AI phone agents are quietly costing stores service business when calls go sideways and nobody's checking the recordings.