Almost everyone who emails me can already build a working agent in Vapi or Retell. Almost none of them have a paying client. The bottleneck was never the tech. It's that you can build the thing and you can't get someone on the phone to buy it.
Here's the exact path I'd run if I were starting today with zero portfolio and zero money.
Pick one boring niche and one city
Not "local businesses." One trade, one metro. Plumbers in Phoenix. Dental offices in Tampa. Restoration in Denver. You want a list where every prospect has the same pain, so your opener and your demo don't change from call to call. Reps compound when the script stays still.
The niche needs three things: it runs on inbound phone calls, each customer is worth $500+, and the owner can't or won't fix the phone problem himself. That's the whole filter.
Get the leads for free
You do not need Apollo or a paid list. Open Google Maps, search "emergency plumber [city]," and start scrolling. Every result has a name, a phone number, and usually a website. Pull 50 of them into a sheet. That's a week of outreach and it cost you nothing but an hour.
While you're there, call each one after hours and see who doesn't pick up. The ones that send you to voicemail at 7pm are your warmest leads before you've said a word. They are literally demonstrating the problem you fix.
Build ONE demo agent, not fifty
Build a single after-hours receptionist for the niche. Generic greeting, books a callback, captures name and job type. Record a 40-second call with it. That recording is your portfolio. You don't need ten case studies. You need one agent that answers the phone like a competent human and a screen recording of it working.
This is the Boiling Frog move: you're not selling the 24/7 AI receptionist yet. You're selling a $500 after-hours agent that stops them bleeding the calls they already miss.
The opener that books the meeting
Do not open with "hi my name is." Open as a customer would:
"Quick question, is this the number I'd call if I had a pipe burst tonight?"
They answer. Then: "Got it. I actually called at 7 last night and got voicemail. I help [plumbers] here in [city] catch those after-hours calls so they don't turn into a competitor's job. Worth 10 minutes to show you what it looks like?"
That's it. You led with their problem, you proved it (you actually called), and you ended on a question. Whoever asks the questions runs the call.
Book, don't pitch
On a cold call your only job is the next meeting, not the sale. The second they show a flicker of interest, stop talking and book the time. "Are you around Thursday at 2, or is mornings better?" Send a calendar invite while you're still on the phone.
Then on the demo, you play the recording, you run their missed-call math live, and you quote the $500 after-hours agent. First client done.
Send 20 of these a day for a week and you will book meetings. The people who don't get clients almost never sent 100 openers. They sent 6 and decided outreach doesn't work.
If you want the openers pressure-tested before you dial, bring them to the Wednesday live. Reading a bad opener out loud in front of the room fixes it faster than any doc can.