The Ledger · Selling
No funnel, no $997 program. Where the work actually comes from when you sell AI builds, and what to say when you find it.
Most advice on getting AI clients is a person selling you a course on getting AI clients. The actual answer is duller and free: you already know where your first client is, you just have not asked them yet.
Start where you already are #
Your first client is almost never a stranger from the internet. It is the small business a friend runs, the team at your day job drowning in a manual task, the person in a group chat complaining about something AI happens to be good at. Warm beats cold every time, because the hard part of a first sale is trust, and you already have some with people who know you.
Make a short list of every person or business you have a real connection to. Next to each, write the one repetitive, annoying, expensive thing they do that a model could help with. That list is your pipeline. It is more qualified than any cold outreach you could run.
Sell a result, not “AI” #
Nobody wants to buy AI. They want the thing on the other side of it: fewer hours on data entry, faster replies to customers, a first draft instead of a blank page. “I build AI solutions” means nothing to them. “I can turn the support emails you answer by hand into draft replies your team just approves” means something.
So lead with their problem in their words, and keep the tech invisible. The model is how you do the job. It is not the job.
The first message #
Keep it short, specific, and free of hype. Something like:
Hey, I noticed your team spends a lot of time on [the manual thing]. I have been building tools that handle exactly that, and I think I could cut it down a lot. Want me to take a look and show you what is possible? No charge for the look.
That last line does the heavy lifting. A free, low risk first look gets you in the door, lets you see the real problem, and gives you something concrete to price. It is not working for free, it is scoping, and it is the fastest path from “maybe” to a paid build.
Proof beats pitch #
You do not need testimonials yet. You need one thing that visibly works. Build a small, real version of the thing for one person on your list, even at a low price or as a favor, and get it actually running. A working demo of their own workflow, with their own data, closes better than any deck.
Then the second client is easier, because now you can say “I did this for them” and point at something real. That is the whole loop: a warm lead, a real result, a number you can show. Repeat it three times and you are not looking for your first client anymore, you are choosing between the ones who found you.
The people selling the $997 program skipped all of this and went straight to selling you the dream. Do the boring version. It works, and it is the only one with receipts.
No course-shilling, no "DM me to learn how." If a number can't be shown, we don't print it.