Your First AI Client in 30 Days

Pick the Pain

Your job for the first three days is not to pick a niche you love. It's to find one boring niche that pays humans to do work a script could do. Boring niches sign fastest because nobody else is pitching them.

Open Claude Code and make it do the digging:

> claude "which local business niches bleed money on
  manual, repetitive work? Scan job boards, forums, and
  reviews. For each: the task, what it costs them, and
  how visible it is from outside."

✓ dentists: no-shows cost $200 a slot
✓ realtors: 5 hrs/wk writing listings
✓ law firms: intake buried in email

Then sanity-check the output yourself. Read ten Google reviews of local clinics. Search "front desk" job listings in your city and note what they're paid to do all day. When the same complaint shows up three times, you've found the pain.

Pick one you can reach this week. This playbook runs with dental no-shows as the example. Swap in whatever your research surfaced.

One Offer

Do not build a services menu. Menus force the buyer to think, and thinking stalls deals. Turn the one pain into a single flat-priced system with one bold promise, and write it down before you build anything.

> cat offer.md
# The Offer
No-show killer for dental clinics.
Every booking confirmed and reminded, automatically.

Price:   $1,000 setup + $400/mo
Promise: cut no-shows in half
Scope:   one workflow, nothing custom

Three rules make this work. Flat price, so there's no negotiation surface. One promise, stated in their numbers, not your features. Tight scope, so you can actually deliver it in a week without drowning in custom requests.

Notice what's missing: the word "AI". The clinic owner doesn't care how it works. They care that empty chairs stop costing them $200 a slot.

Build It First

Here's the move nobody makes: build the entire system for a fake clinic before anyone pays you. Invent "Sunrise Dental", give it a free Calendly account, and build the real thing against it. The working demo IS your portfolio.

The no-show killer is four pieces, and Claude Code can scaffold all of them:

> claude "build the no-show killer for a demo clinic:
  1. webhook endpoint that catches new Calendly bookings
  2. draft + send a confirmation text on each booking
  3. reminder texts at 24h and 2h before the slot
  4. if they reply R, send a reschedule link
  Use Node, Twilio for SMS, and a simple SQLite store."

✓ webhook catches new Calendly booking
✓ confirmation text drafted and sent
✓ reminders fire 24h + 2h before
✓ reply "R" reschedules automatically

> npm run demo
✓ demo clinic live on localhost:3000

Costs to build this: Calendly free tier, a Twilio trial account for test texts, and your evenings. Book fake appointments on your own phone and watch the whole loop fire: booking in, confirmation out, reminders on time, reschedule on reply. Fix what breaks. Run it again until it doesn't.

Seven days is plenty because the scope is one workflow. That's why Level 2 mattered.

Record Proof

One day, three assets. This is the entire sales kit, and it replaces the portfolio, the case studies, and the credentials you don't have.

> ls proof/
demo-2min.mp4     the money shot
before-after.png  the no-show math
one-pager.pdf     offer + price, one page

The video is a 2-minute screen recording, and the script is fixed:

# 2 MINUTES, NO SLIDES, NO INTRO
0:00  a booking comes in on the demo clinic
0:30  confirmation + reminders fire on a real phone
1:30  "that's a $200 slot that didn't go empty"

Record it with whatever you have: QuickTime, Loom, OBS. Keep your phone in frame when the texts land, because that's the moment it becomes real for the viewer. No feature tour, no code walkthrough. Show their pain getting solved, not your stack.

The before-after image is simple arithmetic on their numbers: slots per week, typical no-show rate, dollars per slot, dollars recovered if you cut it in half. The one-pager is your offer.md from Level 2, made presentable.

Lead List

Forty researched prospects beat four thousand scraped emails. You're sending forty personal messages, not a blast, so every row on this list needs three things: the clinic, the owner's name, and a visible gap you can point at.

> claude -p "find 40 dental clinics in Phoenix.
  For each: clinic name, owner or office manager name,
  email, and one visible gap: no online booking, reviews
  mentioning missed appointments, or no reminder system."

✓ 40 clinics found
✓ 31 owner emails verified
✓ gap noted for each row

Verify what Claude gives you. Open each clinic's site, confirm the name, skim the reviews yourself. A wrong name in the first line kills the message. Fifteen minutes per ten leads is the price of every message landing warm.

The "visible gap" column is your opener. "Saw your booking page doesn't send reminders" is research. "I help businesses with AI" is spam.

Send Until Signed

Two weeks, one loop: five personal emails a day, every day. Each message is three sentences and leans entirely on the proof you built.

# the whole email
Hi [name], saw [the gap] at [clinic].
I built a system that confirms and reminds every
booking automatically. Here's a 2-min video of it
working: [link]
If cutting no-shows in half is worth a look,
happy to show you on a quick call.

No attachment, no pitch deck, no "just following up on my previous email" on day two. Follow up once, four days later, with one new sentence. Then move on. Forty leads at this pace covers your whole list with room to re-touch.

When someone bites, the call is a demo, not a pitch. Share your screen, book a fake appointment, let them watch the texts land. Then say the price from your one-pager and stop talking. The system does the convincing, because you built it before you asked anyone for money.

First yes: install the same workflow on their real calendar. That's your setup fee, and days of work you've already done once.