Blog / Build with AI /

Building for yourself first

If you’re not sure what to build, build the thing you wish existed on your own screen this week.

That is my case for building for yourself first. Not because other people don’t matter, but because you are the one person whose problem you can see up close. You know when it happens, how irritating it is and whether a solution is any good. For everyone else, you’re guessing.

Reason 1: you can judge your own tool

When you’re the user, feedback is instant. You try the tool, it either saves time or it doesn’t, and you know why. There’s no waiting for survey answers, no reading between the lines of polite comments.

That matters a lot when you’re working with AI. An assistant can produce plausible code all day. The question is whether the result actually helps. You’re best placed to answer that when the task is yours.

Reason 2: it keeps you honest about the problem

Ideas for other people have a way of growing vague. “A tool for small businesses to manage their social media” sounds fine and describes nothing. “A sheet that turns my list of blog posts into scheduled posts for the week” is specific, because you needed it on Monday.

Specific problems produce specific tools. You know what the input looks like, what the output must be and what annoys you about the existing options. That makes the prompt better and the result sharper. I wrote about this in how to describe a tool to AI so it builds the right thing.

Reason 3: you’ll keep going

Building takes longer than expected, and the interesting part wears off. The motivation that carries a project through the boring middle (bugs, edge cases, polish) is wanting the result. If you’ll use it every week, you’ll finish it.

Projects built to impress an imaginary market tend to be abandoned the first time they get difficult.

Reason 4: your annoyance is probably shared

You’re not as unusual as you think. If you’re irritated by renaming forty images before upload, so are a lot of people. The common tools in everyday work, such as converters, checkers, trackers and formatters, mostly began as someone scratching their own itch.

That’s also why why simple tools win rings true: a tool made for one real task tends to be clear, because its maker never had to invent a feature list.

The counter-argument: “You are not your user”

This is the strongest objection, and it deserves a straight answer. Designers say it for good reason. You know too much. What’s obvious to you is baffling to a newcomer. Your workflow isn’t everyone’s. Your tolerance for rough edges is higher than most. And building for yourself can produce something that’s brilliant for one person and useless to anyone else.

Fair. Here’s where I think the balance falls:

  • Building for yourself is the right way to start. It gets you a working, honest first version quickly.
  • It’s the wrong place to stop. Once the tool works for you, you need to find out whether it works for anyone else.

The risk is mistaking “I like it” for “people want it”. They aren’t the same, and the gap between them is where many products die quietly.

How to move from “mine” to “theirs”

Once your tool does its job for you, take these steps.

  1. Find three people with the same problem. Not friends who’ll be nice, but people who’d really do the task. Show them the tool.
  2. Watch, don’t explain. Give them a task and see where they stumble. Testing your AI-built tool with real users describes how.
  3. Fix the first-minute problems. Your own tool probably has no onboarding, because you don’t need any. Others do. See onboarding for small tools: the first sixty seconds.
  4. Strip out the personal assumptions. Hard-coded paths, your own file naming, settings that only suit your workflow.
  5. Ask what they’d pay or switch from. A polite “that’s neat” is not demand. A request for the link is.

If they don’t care, that’s useful too. You’ve still got a tool that works for you, and you’ve saved yourself from polishing it for a market that doesn’t exist.

Where building for yourself goes wrong

A few traps to know about:

  • Building something so personal that nobody else can use it. It’s fine, as long as you know that’s what it is.
  • Assuming your skills. You’re comfortable with a command line. Most users aren’t.
  • Over-fitting to your quirks. One person’s perfect workflow is another’s obstacle.
  • Skipping security because “it’s just for me”. If it ever goes public, those shortcuts follow it. Hard-coded keys and unchecked inputs are the usual culprits. Never put API keys in front-end code is worth reading before you share anything.
  • Never finishing the last 20 per cent. A tool that works on your machine is not a product.

What this means if you work on websites

People who run websites for a living hit small repeated annoyances all day: renaming files, checking tags, gathering the same data from the same pages, copying a number from one dashboard into a sheet. Each one is a candidate for a small tool. Turning a repetitive task into a tool: my decision checklist is a handy filter for which ones are worth it.

Start with the one that annoyed you most this week. Build it badly, use it, then improve it. That loop is faster than any market research.

Where I land

Build for yourself first, because it’s the quickest path to a tool that works and a problem you understand. Then test on others before you call it a product. Treat “built for me” as the starting line, not the finish.

What’s the most irritating five-minute task you did this week? Could it be a tool by Sunday?