Insights

The gap between a business idea and a working app

The people who understand a problem best can rarely write code. That gap has become much cheaper to cross.

  • Ravi R. Singh
  • 4 min read

Ideas that never get tested

Think of a tuition centre owner in Petaling Jaya who knows exactly what her waiting list should do. Think of a workshop manager who knows what a service reminder should say and when. Neither can write code, and neither will pay an agency a five-figure sum to find out whether the idea works. So the idea stays in a notebook.

This is common across Malaysian business. The Department of Statistics Malaysia reported that in 2024, 95.3% of establishments used the internet and 74.4% had a web presence. That leaves about one establishment in four without a web presence of any kind, and far fewer have a tool built for their own customers.

Imagination was never the barrier. The cost of the first version was.

What changed

AI code generation has cut that cost sharply. A person can now describe a small web app in plain language and get a working version to click through in minutes. No developer is needed, and no specification document.

The important word is small. These tools are good at a single-page app with one job: a quiz that matches a buyer to a product, a menu with an order button, a calculator or a sign-up form. They do not replace an engineering team.

I have seen many ideas die in the brief stage over the years, waiting for budget approval that never came. A first version that costs a few dollars removes that wait. The idea can be judged by the people it was meant for.

Know the limits before you start

Being clear about the limits saves disappointment. A generated single-page app has no server behind it. It does not store customer records in a database, manage user logins or take card payments by itself. It is not built to meet the standards required for financial transactions, medical records or sensitive personal data.

For a first test this is often fine. A waiting-list page can pass enquiries to an email address. A product quiz needs no database at all. The purpose of the first version is to learn whether anyone wants the thing.

Ownership matters too. If you cannot download the code and host it on an account you control, you have rented a demo. Insist on the source files.

A cheap way to test an idea

The method I recommend has four steps. Write the idea as one sentence that names the user and the single job the app does. Build the smallest version that does that job. Put it in front of five real people who have the problem and watch them use it without helping. Then decide whether to improve it, change it or drop it.

Most ideas change at the third step. People ignore the feature you were proudest of and ask for something you had not considered. Finding that out for a few dollars in an afternoon is the whole point.

Be sober about money. A working app is not a business, and no tool can promise income. What it gives you is evidence, sooner and cheaper than before.

What to build first

Good first projects share a shape. They serve people you can already reach, they do one thing and the result is visible on one screen. A bakery owner might build a cake-order form that sends the details to her email. A freelance trainer might build a short quiz that recommends one of three courses. A small distributor might build a price calculator for bulk orders.

Poor first projects also share a shape. A marketplace with buyers and sellers needs accounts, payments and trust on both sides. A booking system for a clinic touches health information. An app that depends on data from someone else's platform can break when that platform changes its rules. Leave these until a simpler idea has taught you how your users behave.

When the test works

If five people use the first version and three ask when they can have it, you have a decision to make. Some ideas can stay as a small tool that supports an existing business, and that is a good outcome. Others need accounts, payments or stored data. At that point a developer is worth paying, because you now know what to build.

The first version has done its job once it makes the next spending decision an informed one.

From notebook to preview

Mazentic is the product we built for the tool choice that comes next. Its Maze Stack is a written report for a small marketing team. You answer a short intake and get a report on which AI tools to use for your workflows, what they cost each month for your team size, and which tools you pay for today to keep or cut. It costs US$79 once.

Other tools exist, and the method above works with any of them. What matters is that the first version is small, the code is yours and five real people see it.

The idea in the notebook has been waiting on a cost that has mostly gone. Test it this month and let the reaction of real users decide what happens next.

Sources

See the four productsAll articles