Worked example: 7-day plan
One fully worked example — from zero to your first customer.
Goal: one complete, follow-along example that ties all the modules together. Copy-paste
messages, prompts, and a day-by-day plan. Adapt it to your niche and get going.
This is no longer theory. We take one concrete little business and walk it all the way through: from idea to the first paid invoice. The example: SEO product descriptions for furniture web shops. The same pattern works for hundreds of niches — swap "furniture" for your market.
The offer (modules 02 & 08)
I help furniture web shops that don't have time to write good product copy, by
delivering compelling, SEO-friendly descriptions with an AI agent — ready to publish.
Price: €99 per month for 50 descriptions (or €2 each to start).
The numbers (know your margin)
Sale: 50 descriptions × €2 = €100.00
- API cost (~€0.02 per description) = € 1.00
- Other (email, hosting) = € 4.00
─────────────────────────────────────────────
Gross margin per package = €95.00 (95%)
Healthy. Your time goes into quality control and customer contact — not into the writing.
Day 1 — Validate before you build (module 02)
Don't build anything yet. First find 5 potential customers and gauge interest. Send this message (LinkedIn, email, or a web-shop owners' group). Copy and adapt:
Subject: quick question about your product copy
Hi [name], I came across your shop [name] — nice collection. Quick question: does writing
product descriptions take you a lot of time? I help furniture web shops with SEO-friendly
descriptions, ready to publish, for a fixed monthly fee. Would that interest you, and what
would it be worth to you? No sales pitch — I just want to know if this is a real problem.
Goal for day 1: 3–5 replies and at least one "yes, I'd pay for that." Listen to the objections — you'll use them tomorrow.
Day 2 — Design the agent (module 04 + the Claude.ai route)
Open claude.ai and refine your instruction together. This is the system prompt — copy it straight into agent_mvp.py:
You are an experienced e-commerce copywriter specializing in furniture.
You write compelling, honest product copy that converts.
Rules:
- Tone: warm, expert, to the point. No exaggeration, no empty superlatives.
- Never invent specifications (dimensions, materials) that aren't in the brief.
- Always deliver: a catchy title, a flowing 2-4 sentence description,
3 to 5 selling bullets, and 3 to 6 SEO keywords.
Test it in claude.ai with a real product from your prospect. Good enough? On to tomorrow.
Day 3 — Deliver once, by hand (module 04)
Use agent_mvp.py (from the code/ folder) with the prompt above. Fill the WERKLIJST (work list) with 5 real products from your prospect and run:
python code/agent_mvp.py
Check each text by hand, polish where needed, and put them in a clean Google Doc or spreadsheet. This is your free sample delivery — quality matters more than speed here.
Day 4 — The sales conversation (module 08)
Send your sample to the prospect who was most interested. Copy and adapt:
Hi [name], as a taster I went ahead and made 5 product descriptions for you — see the
attachment. Feel free to publish them and see what they do. If you like it, I deliver 50 of
these every month for €99. The first month is no-obligation. Shall I get you set up this week?
Three common objections (and your reply)
| Objection | Your reply |
|---|---|
| "Isn't it just AI?" | "Exactly — that's why it's fast and affordable. But I check every text myself, so you get AI speed with human quality." |
| "€99 is a lot" | "How many hours does it cost you now? For 50 texts I easily save you a full workday a month. Convert that to your hourly rate." |
| "I want to see it work first" | "Totally fair — that's why you already have 5 for free. Publish them, and only pay from the next batch." |
Day 5 — Set up payment (module 09)
Keep it simple for customer number one:
- Send a clean invoice (or a Stripe Payment Link for €99).
- Set VAT aside (module 11).
- Record income and costs in a spreadsheet from euro one.
Automated payments and webhooks come later — first prove that people pay.
Day 6 — Deliver and guardrails (modules 06 & 10)
Deliver the first real batch of 50 descriptions. Turn on your guardrails when you run it:
- a daily budget in
agent_mvp.py(already built in), - a spending limit in the Anthropic console,
- and check a sample of the output before you send it.
After delivery, ask for feedback and a short testimonial — you'll use it for customer two.
Day 7 — Repeat and automate (module 12)
Now that you know it works:
- Reach out to 5 new furniture web shops with the same day-1 message.
- Write down every manual step from this week — that's your automation list.
- Automate the most tedious step first (e.g. drop the texts into a Google Sheet automatically via
Zapier, module 09).
Repeat the cycle. Three to five customers in the same niche is a real, running little business.
The 7-day checklist
- Day 1 — reached out to 5 prospects, at least 1 "yes, I'd pay for that"
- Day 2 — system prompt designed and tested in claude.ai
- Day 3 — delivered 5 texts by hand as a free sample
- Day 4 — sent the sales message, handled objections
- Day 5 — payment method set up (invoice or Payment Link)
- Day 6 — delivered first paid batch, guardrails on, asked for feedback
- Day 7 — reached out to 5 new prospects, automated the first step
Handy tools (concrete)
| Function | Start with |
|---|---|
| AI engine | Claude API (claude-opus-4-8), or claude.ai to design |
| Running code | Python on your own computer (module 03) |
| Payments | Manual invoice, Stripe Payment Link, or Gumroad |
| Email/delivery | Google Docs/Sheets, a transactional email service |
| Integrations | Zapier (9,000+ apps) — start with 1 trigger + 1 action |
| Bookkeeping | A simple spreadsheet, later a bookkeeping tool |
The core idea
This example works because it's narrow (one service, one niche), validated before building, and delivered semi-manually until you know people pay. Swap the niche, keep the pattern, and you have a playbook you can run again every week.
Start today at day 1. One message actually sent is worth more than a week of perfecting.