The case at a glance
The operator ships an AI post-call workflow tool for solo consultants. Zoom in, branded recap email and draft proposal out, before the coffee cools. Not another note-taker. A deliverable machine.
- Total build time: 72 hours across 4 days
- Time to first paying user: 9 days
- Month 1 revenue: $6,092
- Subscribers at day 30: 62 at $79/mo
- One-time "meeting-to-proposal" packs sold: 6 at $199
- Free trial → paid conversion: 41%
- Direct API cost per subscriber/mo: $4.10 (94.8% gross margin)
Stack. Next.js 15, Supabase, Stripe, Claude 4.5 Sonnet, Recall.ai, Resend, Vercel, Loops.so for onboarding. Total infra under $80/mo. The operator is a scarred solo builder on his 21st attempt. The first 20 projects, combined, never cleared $6K/mo. This one did in four weeks.
What he's actually selling (and why most AI meeting tools miss it)
The product isn't the transcript. The product is the 90 minutes between the call ending and the client email being sent.
A thousand tools sell "AI meeting notes" for $10/mo. Fireflies. Otter. Granola. Read. That layer is a commodity and it's priced like one. Project 21 skipped the layer entirely and sold the deliverable instead. Branded recap email in the consultant's voice. Draft SOW pulled from the scope talk. Action items pushed to Asana. One click to send.
A consultant billing $250/hr doesn't do the math on $79/mo. They just click. That's the whole trick. The commoditized layer is priced against; the un-commoditized layer is priced with.
Every profitable AI wrapper in 2026 has quietly done the same move — one step downstream, from summary to deliverable, from token to hour. The operators still competing at the summary layer are running 40% margins and wondering why paid ads never pay back.
Sell the recovered hour, not the model.