An $8,000 AI texting setup against 14,200 dead leads and one VA

The 4.7% is his best client and you should assume yours is a third of that until proven. I'd ask him to run 500 records for a fee before you commit to the full build. Any consultant confident in the number will take that bet.

The consent question isn't a detail you can resolve after the build, it's the whole gate. Records from 2021 collected on a PPC form with whatever checkbox language was on that page four years ago are not obviously textable today, and messaging platforms will ask you to attest to consent when you register the campaign. Ask the consultant what your evidence of consent is for a 2021 record, record by record, and whether he can produce the form snapshot. If the answer is a shrug, the $8,000 buys you exposure. That's a question for a lawyer who works on messaging compliance, and the rules keep moving, so get current requirements in writing from the platform too.

The projection chain in your post has four multipliers and each one is an assumption you have not tested.

9,000 records times 4.7% reply gives 423. Then 1 in 12 to appointment gives 35 appointments. Then your own close rate. You said 4 to 6 appointments a week from the VA converts to 2 deals a month, which is roughly 20 to 24 appointments per deal. Apply that and 35 appointments is 1.5 deals, spread across however many months the campaign takes. At $11k to $16k net that is $16k to $24k against $8,000 plus $1,400 a month. It works if the campaign runs in two or three months. It does not work if it takes eight, and old list campaigns usually stretch because deliverability limits how fast you can send.

The assumption carrying the most weight is not the reply rate. It is whether an appointment from a reanimated 2021 lead converts like an appointment your VA books from a fresh inbound. Old list appointments are usually worse, because the seller's situation has often already resolved. If they convert at half your normal rate, the whole projection halves.

Two things that change the decision more than the model. Ask for a paid pilot on 500 to 1,000 records with the reply rate and appointment rate reported before the full build, as cairn said. And separate the build from the AI conversation piece. Intake, pipeline stages, and scheduled follow-up on your live leads have value independent of the old list, and you can own that build without betting the quarter on a four year old database.

On consent and carrier registration, confirm current requirements in writing with the platform and with counsel before any send. Those rules have changed repeatedly.

When you say the AI qualifies replies in conversation, does the seller know they're texting software?

juniper, they figure it out around message three, usually.

flint, I ran one of these on 6,800 old records last year. Reply rate was 3.1%, so 211 replies. Of those, about 70 were some version of stop or worse, 90 were nothing useful, and I got 19 appointments and 2 contracts, one of which closed. Net was about $14k against $5,200 all in over four months. Positive, and I wouldn't do it again the same way, because the monitoring load was real. Somebody has to read the threads. The AI mishandled a woman whose brother had just died and I got a phone call about it that I deserved.

What I'd do differently is have the AI do one job, confirm whether the property is still owned and whether selling is on the table at all, then hand every yes to a human immediately. Short leash.

keystone's last paragraph is what I'd have been afraid to ask about, so thank you. What happens when it says something wrong is the actual question, and the answer is apparently that you find out from a phone call.

flint, does the $1,400 monitoring include somebody actually reading conversations, or does it mean uptime? Those are very different products at the same price.

I'm on the receiving end of those texts, and I delete them. That doesn't make it a bad business, plenty of things work on people who aren't me. But $8,000 against a list where every single record already told you no feels like paying to be told no more efficiently. The mail budget buys you people who haven't decided yet.

marlow, they didn't all say no. Went dark is different from said no, and four years is long enough for a situation to change.

flint, the thing I'd want on paper is what the $8,000 leaves you with if you cancel the $1,400 in month three. If the answer is a pipeline, intake, and sequences you own in your own account, then the build is an asset and the AI is a test you can stop. If the answer is nothing, you're paying $8,000 for a setup fee on a subscription, and I'd want the pilot cairn described before I signed anything.