AI voice on the first touch: does it cost you the seller who would have talked?

The unmeasurable hangup is real and I don't think it's small. I've sold two parcels in my life and both times the buyer who got me was the one who sounded like he'd driven past the property.

@orchard but you're one seller describing what worked on you, and Wren's friend has thousands of attempts. I keep getting stuck at exactly this, the anecdote feels truer and the data is what I should use. I voted AI first with human handoff and I'm not confident.

It may help to separate three things that get lumped together as "AI calling," because the poll options mean different things depending on which you have.

A predictive or power dialer is software that dials many numbers at once and connects a live human caller when someone picks up. The human does all the talking. This has been standard for years.

A voice AI agent actually speaks. It delivers an opening, handles a few responses, and either books something or passes the call to a person. Quality varies a lot between systems, and so does how quickly it hands off when the conversation goes sideways.

AI in the back office does neither of those. It scores which records to call, drafts follow-up, and summarizes recordings. Nobody on the other end of the phone experiences it.

For a beginner reading this thread, the third kind is the least controversial and the first kind is the most established. The middle one is where the disagreement here sits.

On Wren's disclosure point: rules governing automated calls, prerecorded messages, consent, and do-not-call obligations exist at both federal and state level in the US, they differ by state, and enforcement and the rules themselves have been changing. Anyone about to run an automated calling program should have a lawyer review the specific setup and confirm current requirements in writing rather than relying on what a vendor says on a sales call. That applies whether you run it yourself or hire it out, because the investor whose name is on the campaign is usually part of the exposure.

Ran voice AI on 12,000 records over about ten weeks. Booked appointments fine. The problem showed up at the appointment, where roughly a third of the people had no memory of agreeing to anything and one told me flat out he thought he'd been talking to a person and felt tricked. Show rate was about 20 points below what my human caller produces. So my cost per appointment looked great and my cost per contract didn't move.

@flint that's the most useful number in this thread. Cost per appointment is a vanity metric if show rate moves against you.

Asking the embarrassing one: when the AI hands off to a human, what does the human hear? Do they get the recording, a transcript, nothing? Because if the seller has to repeat their whole situation the handoff is where you lose them, regardless of who dialed.

@compass on the systems I've seen demoed the human gets a summary in the CRM and the recording is available. Whether the caller reads either one before picking up is a management problem, not a technology problem.

I'm mostly lost in this thread but the felt tricked comment stuck with me. If someone signs a contract after that, is there a story where it comes back on you later? Probably a lawyer question, ignore me if it's silly.

@vellum it isn't silly, and it's exactly why I'd want the disclosure question answered in writing before running the program rather than after the first contract. Not something anyone here should answer for you.

Voted human always and I know I'll be in the minority. My reasoning is boring. Five times the attempts at the same conversion is five times the annoyed people in a market I have to keep working in.

@verity the market saturation angle is underrated. In a county with 30,000 parcels you can burn the whole owner base in two years and then nothing works, human or otherwise.

Small point from the ownership side. I get these calls now as an owner of eleven units, and I can tell within four words. The AI ones I hang up on. The humans I sometimes talk to out of professional curiosity, and twice that turned into a real conversation about a building I'd actually consider selling. Sample of one landlord, do what you want with it.

@cairn but you're a sophisticated owner. The tired inherited-property owner in another state who hasn't thought about the house in four years probably can't tell and probably doesn't care. Maybe the right split is by list type rather than by first touch generally, which is what option three is saying.

Option three got my vote for that reason. Cold bulk is a filter, warm is a conversation, and using the same tool for both seems obviously wrong once you say it out loud.

I'll disagree with the whole framing. My constraint isn't first touch quality, it's follow-up capacity. I have 400 people who said maybe in the last year and I contact maybe 60 of them a month. Whatever tool gets me through that backlog is the one I want, and if the AI is better used on touch four than touch one, that's where I'd put it. Flint, did you try it on your existing maybe list rather than on cold records?

@keystone that reframing is the one I'd want tested. The cold list is where a bad first impression costs the most and the warm backlog is where nothing is currently happening at all, so the downside of a mediocre AI touch on someone who already said maybe is much smaller. Wren, worth asking your friend's vendor whether they've measured it that way, because I'd bet they haven't.