The Non-Coder CEO's AI Stack

This page — and the entire platform it sits on — was built by one person with no developer background: 370+ commits, 64,000 lines of code and more than 70 pages in six weeks. Here are the exact tools, the process that comes before them, and the things you should not use AI for. No course, no signup — just the recipe.

The tools

Seven tools. The first five require no technical background to understand — the AI is the one operating them.

Claude Code — builds

Writes the code, fixes its own mistakes and explains what it is doing in plain language as it goes. You describe what you want; it builds it. It is the only "developer" behind this platform.

Codex — checks

A second AI reviews every single change before it goes live, like a critical senior developer. Two AIs watching each other catch the mistakes you cannot see yourself.

GitHub — remembers

Stores every version of everything. If something breaks, you roll back to yesterday. It is the safety net that makes letting an AI build feel safe.

Netlify — publishes

Puts the site on the internet automatically every time a change is approved. No servers, no IT department.

Supabase — stores

The database. Your business data, stored safely — the kind of thing you used to hire people to run.

Stripe — charges

Payments and subscriptions. Connect it, and billing runs.

A file-based "second brain"

Ordinary folders with research, decisions and logs that the AI reads before it works. It is the difference between an assistant that knows your business and one that guesses.

The process before the AI

The tools above are the easy half. The hard part is the same as it was before AI existed: knowing precisely what should be built, and why.

AI builds what you describe. If the description is unclear, it builds guesswork at high speed and with great confidence. Every piece of work behind this platform therefore started in the same place everything else in a business starts: with the processes.

Before opening the AI, I write one page in plain language answering six questions:

Which process should this even be about? That is decided with two lists:
- What works today and could become faster or cheaper? With or without AI: some gains only require better organisation.
- What does not work today and might start working with AI?
What should it do?
Who is it for?
What would make it wrong?
What is deliberately left out?
What proves that it works?

If the page cannot be written, the problem is the process. That is where the work is, and the AI has to wait. It is also why years of experience with operations and processes are the real head start for an SME director: you have filled in that page a thousand times before, just with different headings.

Once the page is written, the rest is instruction. Ask the AI to question the page, one question at a time, until it can lay out a plan. Let it build step by step and explain as it goes in plain language. And when it says done: demand the proof. What was tested, and what was the result?

The free starting point: the prompt for question 1

Question 1 is the only one of the six you can let the AI help you with before you have decided anything at all. Copy the prompt below into Claude or ChatGPT and set aside 30-45 minutes. Answer with real numbers, not gut feelings. I run it in Claude Code myself, the same tool that built this platform. The result is the two lists and one recommendation you can stand behind.

You are my sparring partner on processes. Technology comes later. I am an SME director in a company with [number] employees in [industry]. Your job is to help me map my processes before we talk about AI at all. How you work: 1. Ask me one question at a time, and wait for my answer before asking the next. Ask at most twelve questions in total; then summarise what you have. 2. Start by mapping the processes that take the most time: what do we do every day or every week, who does it, how long does it take, and what happens when it fails? 3. When you know enough, split the processes into two lists. List A: processes that work today but could become faster or cheaper. For each one, mark whether the gain requires AI or could be achieved with better organisation alone. List B: processes that do not work today, where AI might make a difference. For each one, mark what is missing today: time, knowledge, hands or a decision. 4. Recommend ONE process to start with, and justify the choice with my own answers, never with general arguments for AI. 5. Finish by telling me what you do NOT know about my business that could overturn the recommendation. Rules: - No tool recommendations before the mapping is done. - No enthusiasm. If a process is healthy and the gain is small, say so and leave it alone. - Never guess. If you are missing a number or a fact, ask me or write "I don't know". "I don't know" is a valid answer. - Every conclusion must point to something I actually said, quoting my own words. If it cannot, it is a guess and has to go. - When you summarise my answers, use my numbers as I said them. Never make your own fractions, averages or conversions of them.

Note the last three rules. They are the difference between a sparring partner and a salesman: it must never guess, it must quote your own answers when it concludes, and it must be allowed to tell you that AI is not worth the effort here.

What not to use AI for

This section is missing from most AI guides. It is the expensive half:

Facts you have not verified yourself. AI sounds most confident precisely when it is wrong. Demand sources and proof — every time.
Anything published without a human in between. The AI builds; you approve. Always in that order.
Your judgement calls. Pricing, people, strategy. The AI can lay out the groundwork — the choice is yours.
Selling. AI builds the product — it does not bring the customers. The daily conversations, the trust, the relationships — that part is still yours. Plan for it from day one.

The proof

Everything you see on boardready.dk — every page, every feature, every animation — was built with the tools and the process above. Not as a demo, but as a real platform with login, payments and live data.

Questions?

Happy to share what worked and what did not. Message me on LinkedIn — I answer myself.

Message Kenneth on LinkedIn