🧾 I Built A £47 Product With Claude (Here's The Exact Process)

There’s a genre of internet post that goes “I made a digital product with AI in a weekend” and then gets vague about everything that matters. This is the non-vague version. The Claude Kickstart is my £47 guide, it’s on this site, people buy it, and Claude helped build nearly every part of it.

The honest headline first: Claude did the typing. I did the knowing. The product is six years of doing this work for clients, compressed. Claude made the compressing take weeks instead of months, and it could not have supplied the six years. Anyone selling you “AI writes your product for you” is selling you a PDF nobody finishes.

Step 1: Sell something you’re already asked about

Not “brainstorm product ideas with AI”. The opposite. Your product is the thing people already ask you: the question that keeps arriving in your DMs, the thing clients pay you to do that some of them would rather learn. Mine was “how do I actually get set up with Claude properly?”, the question I kept being asked. If nobody’s asked you the question yet, you don’t have a product yet; you have a guess.

Step 2: Interview yourself before you draft anything

The single highest-value prompt of the whole build:

Nick this one:

You're interviewing me so I can turn what I already know into a paid guide. I've answered this question for clients a hundred times. Your job is to get it out of my head in usable detail.

Topic: [e.g. "how to actually get set up with Claude properly"]
Who it's for: [e.g. "self-employed people who've opened Claude twice and closed it again"]

Interview me one question at a time, waiting for each answer, and cover:
1. What they're struggling with before they find me.
2. What I do differently from the standard advice.
3. My actual process, step by step, including the bits I do without thinking.
4. The mistakes I watch people make.
5. The questions I get asked every single time.

Rules: one question at a time, never a list of five. Push for specifics. If I answer vaguely, tell me so and ask again. Don't summarise my answers back to me, don't compliment them, and don't start writing the product. Keep your own words short, I'm the one who should be talking.

When you've got enough to outline the whole thing, say so and stop. Then give me my answers back as a document I can download, tidied only for the "ums". That document is the source material for everything else.

Before you begin, ask me up to three questions about the product and who's buying it, so your interview lands.

I answered in voice mode over a couple of sessions, rambling like I would to a friend. That transcript, with my phrasing, my examples and my opinions, became the source material for everything. This is the step people skip, and it’s why their AI-built products all read like the same product.

Step 3: Draft in pieces, never “write my ebook”

One section at a time, always pointed back at the interview.

Then this, once per section:

Draft one section of the guide: [section name]. Work only from my interview answers.

Give me:
1. The section itself, in my words, keeping my phrasing and my examples even where they're rough.
2. A list at the end of anything you added that I didn't actually say, so I can cut it.
3. Nothing else. No introduction, no summary, no "hope this helps".

Rules: English spelling, short sentences. If my answers don't cover something the section needs, leave a gap marked [GAP: ask me] instead of filling it in yourself. Never invent an example, a number or a client story.

Ask me anything you need to know first.

Then the bit that actually took the time: reading every draft and marking where it stopped sounding like me. On a long build the drafts drift. They get smoother, more reasonable, more like everyone else’s. I got one back and thought, this is slop, this isn’t me. The fix was my own red pen, putting the rough edges back in, and telling Claude which exact sentences were the fake-sounding ones so it stopped producing them. Budget real time for that pass, and if you want the shortcut, it’s a guide of its own: Stop AI Writing Like AI.

Step 4: Let Claude do the boring production

The parts that used to need a designer or a lost weekend: formatting the document into a branded PDF, the sales page copy (drafted, then de-slopped, same red pen), the product description, the launch emails. Claude did the scaffolding. I did the deciding.

Step 5: Sell it somewhere boring

Mine’s on Payhip: free to start, takes a small cut per sale, handles VAT & delivery, took an evening to set up. Gumroad & Lemon Squeezy do the same job. Do not build a website first. The product plus a checkout link is a business; the website can come when it’s earned.

The honest bit

The whole thing took me weeks of real evenings, not a weekend. Anyone quoting a weekend is not counting the red pen. £47 products don’t make anyone rich at small-audience scale either; mine exists to help people properly and to be the front door to my bigger work, which is the honest maths for most of us.

And the uncomfortable one: AI has made average products easy to make, so nobody pays for average any more. The only thing left is that yours couldn’t have been written by anyone else. Protect that above everything.

The one to do first

Run the self-interview prompt tonight, voice mode, twenty minutes on the question people already ask you. Don’t build anything yet. Read the transcript back and notice how much you already know. Most people never find out they were carrying a product around.

That’s the skeleton out of your head in one evening. The Claude Kickstart is the product I built this way, and it does the same trick on your side: Claude set up around your actual trade, sounding like you, five real jobs handed over. £47.