# What to build first

## Start from a problem you can see

Say you work in a school office and you spend every Friday copying numbers between two spreadsheets. That's a build. Say your mum runs a market stall and keeps orders in a notebook. That's a build too.

The best first ideas come from places you already know:

- Your job. What do you do every week that's boring and the same each time?
- Your home. What does your family keep forgetting, losing or arguing about?
- A local business. Think of a plumber, a landscaper or a café that has no website, or a bad one.

Write down ten. Don't judge them yet. Ten rough lines in a note on your phone is enough.

## Copy a proven structure, make the identity your own

Don't invent a new layout or a new way of using a thing. If a type of website or app already works, people already know how to use it. Copy how it's laid out and how a visitor moves through it. Then make the look, the name and the voice yours.

Say you want to build a site for a landscaper. Search "landscapers in London", find one that looks good, and use it as a design reference. The structure is proven. The photos and the words are the landscaper's own.

## Pick one and keep it small

If you keep putting this off because you don't know where to start, pick one area and go. It might be a website for a local business, a small automation, or a tool for your own work. You'll learn the same basics from any of them: how a project is laid out, what an API is, how things get put online.

A first build should take about a week. If it would take a month, cut it in half and cut it again. Ask yourself what the smallest working version looks like. One page. One button. One thing that does its job.

Here's where people get caught. You tell the AI your plan, it says it's a great idea, and then it suggests another feature, and another. Before you know it you think you've got the next big thing, and nothing works. Get the first small part working and tested before you add anything.

## Check someone wants it before you build

Before you spend a week, spend an hour looking for demand.

1. Search for what you want to build and see who already sells it.
2. Read the comments and reviews on those products. What do people moan about?
3. Ask three real people: "Would you use this?" Ask what they do today instead.

If something already exists, people probably pay for it. I look at how well they're doing and whether that applies to me. If a piece of software is dead expensive and you wish you could use it, ask Claude to help you work out how it's built and whether you could make a version for yourself.

Check the legal side early too. Data protection and each platform's rules should be sorted on day one, because they're painful to fix later.

## Does it have to make money?

No. If you're brand new, you need experience with the technology first. If everything has to earn, the hard part becomes marketing and sales, and you end up with something nobody uses. Put a small thing live for free, get a few real users, and you'll learn about running something in production.

## Prompt to try

Open a project in Claude or ChatGPT, start a chat, and paste this. Fill in the brackets.

```
I want to build my first project with AI and I keep procrastinating.
Here's my job: [what you do]. Here's what annoys me about it: [list].
Here are a few other ideas: [list].

Pick the three that could be finished in a week by a complete beginner.
For each one, tell me who else already does it, what people complain
about in reviews, and the smallest version I could build first.
Then ask me questions until you're sure which one to start with.
Don't invent facts. If you're not sure, say so.
```

Then ask it to teach you as you go.

**Next:** [Tech to watch](tech-to-watch.md)
