I ran an AI agency for 18 months. Custom funnels. Retainer clients. The whole infrastructure.
Discovery calls all day. The pain wasn't the offer — it was the lead stream. 70% of every booked call was junk. Wrong-fit budgets, wrong-niche curiosity, "let me think about it" tire-kickers. To scale this I hired a 20-person sales team. Setters on the front, closers on the back.
Fix-cost reality check (Glassdoor median, US 2025):
closer base salary alone = $36,000 / year. Twenty seats × $36k = $720,000 / year in W-2 base before any commission, before SDR salaries, before benefits + tools + management. $100k+/mo just to keep the funnel breathing — and most of those calls still went nowhere.
I was running a sales org instead of a product company. Margin disappeared into headcount. The lifestyle became "manage humans who manage humans".
The pivot was two moves.
- — Wired an AI agent in front of the calendar. It qualifies every inbound lead in 3–5 DMs in my voice. I read its report on each lead before deciding whether to book. Most of the junk calls just stopped showing up.
- — Repackaged the system as a $200 self-serve constructor — the entry tier. Self-serve, full system, sufficient to ship the whole funnel solo. The done-for-you upgrade stayed available for buyers who later asked for it.
I expected the math to be simple: most of the sales-team payroll gone, same revenue staying on the table (the few qualified leads were the ones actually closing anyway). That part happened — and margin lifted hard. The unexpected part is what came next.
The buyers started DM-ing me.
A few weeks after picking up the $200 constructor, the same people — entry buyers I'd never spoken to — would slide into my DMs asking for a call. Not the desperate "convince me" kind. The "I'm ready, let's talk about the install" kind. Almost every one of those calls closed at $5–20k.
It took me a while to see why. Two reasons:
- 1. The constructor pre-sold the value. They opened it, used the tools, ran the prompts. The "is this BS" objection closed itself before they ever pinged me. By DM time, they already knew the method worked. The conversation wasn't about convincing — it was about choosing whether to install it solo or with a mentor.
- 2. They were paying for speed. Once they understood the system, the only open question was "how fast can I run this end-to-end". They weren't paying $5–20k for content. They were paying to skip weeks of solo grinding and have me install it with them.
70% of cold-call hours — gone. Net income — 5×. Different leads. Different motion. Same product — just sold in the right order.
That's the model wired into printer/ai. Same two moves, productized.