Pretty much every dealership phone system records calls now. Almost nobody listens to them. I once asked a GM how many recordings his store had sitting on the server and he laughed and said "all of them," which was the honest answer, since a manager pulls one a week, give or take, when somebody complains and the rest just pile up. Hours and hours of customers explaining exactly what they want, and the store never hears a word of it. Turn those calls into text and things get a lot more useful. You can search text. You can skim forty calls over a cup of coffee. You can count how many times people said "loaner" in October. Audio you have to sit through at 1x speed, and nobody's got time for that.

What callers keep asking (it's not what's on your website)

Read a week of sales calls and you'll notice something kind of funny. All that stuff on the vehicle detail pages, trim walkarounds, feature lists, the slick 360 photos? Barely comes up. People call to ask four or five things, over and over. Is this specific car actually on the lot right now. What's the real out-the-door number once you add fees. Does that advertised price assume a rebate I don't qualify for. Can somebody give me a rough number on my trade without me driving forty minutes. Two things fall out of that. Your salespeople should have straight answers ready for those exact questions, because callers can smell a dodge from a mile away, and a dodge sounds like "they're hiding something." And your listings could probably answer half of them up front, which means fewer calls from people who only wanted to know if the car exists.

Where calls go to die

Put transcripts next to CRM outcomes and the dead calls start to rhyme. "Let me check and call you back" shows up constantly, and the callback either never happens or happens the next afternoon after they've bought somewhere else. Two-minute holds while somebody jogs the back row looking for a stock number. Oh, and my personal favorite, the price question answered with "so when can you come in?" Three times. In one call. From the customer's side that's a brick wall. You can't see any of this on a call report. Four minutes, twelve seconds, outbound, connected. Great, thanks.

Service has even more to say

I'd argue fixed ops learns more from transcripts than sales does. Search a month of service calls for "still waiting," "loaner," or "nobody called me" and stuff advisors already half knew suddenly comes with a count attached. You might find a third of status calls coming from people who were promised an update by noon and never got one, or a recall that gets three different answers depending on which advisor picks up. One store I talked to discovered a single backordered part was generating something like a call every other day from one customer, who got a little angrier each time because nobody could give her an ETA. Those are process problems. Cheap ones to fix, compared to the cost of answering that call fifty times.

Free market research, more or less

Transcripts make a messy but surprisingly honest demand survey. If a dozen people this month asked about hybrids under thirty grand, or a three-row anything under 40,000 miles, your used car manager wants that before the next auction, not after. Marketing gets something out of it too. Callers mention where they found you more than you'd expect, the radio spot, a mailer, "that Facebook ad with the guy in the hat." Nobody should treat that as proper attribution, but as a sanity check on where the ad money's going it's better than nothing.

Training with real calls instead of role-play

Role-play has its place, I guess. But sitting a new salesperson down with two real calls about one model, one that turned into a Saturday appointment and one that ended in "I'll think about it," teaches more in ten minutes than a week of scripts. And because you're both looking at the actual words, there's no "that's not how it went" argument.

Getting transcripts without anybody typing

Transcription used to be a chore. Someone listened and typed, or you paid a service and got files back days later that lived in a folder nobody opened. These days the phone answering itself can handle it. An AI front desk for car dealerships sitting on the main line produces a transcript and a short summary of every call it takes, tags it sales or service or parts, and notes what the caller wanted and what happened next. The record exists the second the call ends.

Put them where people already look

Plenty of stores trip up right about here. A pile of transcripts in some separate dashboard is just a fancier version of the recordings nobody listened to. What makes AI call transcription pay off is the notes landing in the CRM, the scheduler, the service calendar, the stuff your team opens every morning anyway. A salesperson calling someone back sees what they asked about Tuesday. An advisor sees this is the customer's third call about the same rattle and maybe skips the "so what seems to be the problem?" opener. Three clicks and a separate login, and nobody will bother. Ask me how I know.

Before you flip it on

Recording has legal strings. Some states require everyone on the call to consent, so your disclosure at the start of the call needs to be right for every state you take calls from. That's a conversation for whoever handles compliance at the store or group, not something to guess at. Transcripts also pick up personal stuff (addresses, credit situations, the occasional partial account number) so keep access tight and decide on retention on purpose. And don't try to read everything. You'll drown. Pick a few questions you care about each month, search for those, read a small sample start to finish, fix what you find. Stores that do that boringly, month after month, end up knowing their customers better than the dealership across the street, and they never mailed out a single survey.