Build · Custom app

Man Cave queue system

A four-location barbershop runs its walk-in floor on rented queue software. I built them a replacement, deployed it as a working demo, and used the demo itself as the sales pitch. The demo played a full shop day on its own and ran for two months while the owner weighed the offer.

The gap it closes

Today: the cut ends, the barber walks to the computer, clicks, and the system texts the next client, who is somewhere in the mall. Cost: 5 to 10 minutes of empty chair per handoff.
With this system: the next client gets texted 10 minutes before the cut ends, so they are walking back while the barber finishes.
The barber's whole day runs on one button on their phone.
The dashboard counts recovered dead-chair time, a before-and-after the owner can measure.

Man Cave queue system demo landing page — The demo site is the pitch: a guided example, an interactive sandbox of every screen, and a plain statement that everything is real and running.

Five screens, one engine: check-in kiosk, barber phone view, client SMS thread, in-shop waitlist TV, owner dashboard.
Fair "first available" rotation goes to the barber idle longest, with the reason printed in plain English, so the rotation is never an argument in the shop.
Two automatic texts per client, a check-in confirmation and the heads-up, and nothing else. Clients in the shop get no texts at all.
An AI lane reads client replies: clear answers run themselves, and anything abnormal gets flagged to staff. The AI never writes to a client.
A decision log records every automatic choice in plain English, so the owner can audit the machine.

How I built it

The engine. Cloudflare Workers, one Durable Object per shop, which makes multi-location and white-label a URL away. n8n behind a webhook as the SMS layer.
The AI lane, fenced on purpose. A small fast model only classifies client replies: confirm, late, cancel, or unsure. Unsure flags a human. A bigger model answers owner questions, grounded strictly in the app's own numbers. The AI never writes to a client. For owners who are cautious about AI, that boundary is the whole pitch.
The self-playing demo. Seeds a simulated Saturday, the busiest traffic curve, and plays the whole day: cuts finish, walk-ins arrive tagged simulated, clients text back C on their own. The numbers build from real events, never a script.
Production reality. The free hosting tier's request cap took the site down one day. I rebuilt the polling model so hidden tabs cost nothing.
The pre-send review. Five perspectives before the email went out: owner, manager, UX critic, sales lens, developer. All nine resulting fixes shipped, down to owner-editable SMS templates so a copy question could never stall the deal.

Man Cave owner dashboard — The owner dashboard: all four locations, live queues, the SMS feed, stats, and the decision log.

What the build taught me: an emailed demo can do the work of three meetings if the demo runs itself and cannot be broken by curiosity. Selling AI to a cautious buyer means designing for auditability, plain-English reasons on every automatic decision, and being honest in the pitch that the dead-chair gap does not happen on every cut. Modest claims that hold up beat big claims that need defending.

Why this matters for selling AI

Founder-led sales against an incumbent subscription, closed toward an in-person pricing meeting with a working product instead of a deck.
Pricing anchored to recovered revenue: one saved cut a day per location is worth roughly $1,000 a month against a much smaller fee.
Transparent standard pricing, with a founding-client offer for the first shops.
Same commercial motion as any high-ticket close: value math the buyer can check, proof they can touch.