Blog
Mockup first, build later: why that matters even more with AI
· 4 min read
You have an idea for an application. In the past, that meant waiting weeks for a first sketch. Now you type a description into an AI tool and within minutes something working appears on your screen: screens, a data model, some logic behind it. The temptation to carry straight on building is strong. Why spend time on a mockup when the real application is almost there?
That temptation is exactly the problem. The faster AI produces a working version, the easier it is to skip the question of whether it is the right version.
Why AI makes it more tempting to skip the mockup
On its product page about Mentor, the platform's AI-driven app generator, OutSystems itself states that you enter a description in natural language and get a working application out of it, including a data model, logic and user interface. The platform also claims that this can cut development time by up to 60 percent (outsystems.com).
That is precisely the point. What used to take a few weeks of build time, and therefore naturally left room to check along the way whether the direction was right, now happens in a few minutes. There is no longer a natural pause in which a client, a colleague or a future user can say: this is not what I meant. On top of that, an AI tool readily confirms that something works without it actually having been tested against a real situation. The result looks finished because the fields are filled in and the buttons respond, but that says nothing about whether the screens, roles and sequence are the right ones.
The less time building takes, the more important it becomes not to let the step before it disappear: working out and testing what exactly needs to be built.
What a mockup still solves that AI does not replace
A mockup is not a delay; it is a cheap way to prevent an expensive mistake. A clickable sketch with realistic sample data, in both phone and desktop formats, shows a client straight away whether the direction is right before any build time goes into it. An employee clicking through a mockup will spot a missing screen or a step in the wrong order much faster than the same person reading a list of features.
Interestingly, OutSystems recognises this principle in its own AI workflow too. According to the platform's documentation, when generating an application with AI, a user first goes through a step in which the blueprint, the AI's interpretation of what was asked for, is reviewed and adjusted before the full application is generated (success.outsystems.com). In other words: even a platform that delivers a working app within minutes builds in a checkpoint before the real work begins. That logic applies just as much, and perhaps even more so, to the conversation with your own client or users.
The second reason a mockup keeps working is that someone else has to be able to reject the result before it is called "done". The builder, human or AI, can no longer see their own assumptions. A colleague or future user who looks at the screen with fresh eyes will catch them. That principle does not change just because the first version comes together faster.
A concrete example from practice
For the certification platform we built for a security company with more than fifty affiliated dealers and hundreds of vehicles a year, the process had previously run entirely through email and manual checks. That led to chaos and delays. For an application that had to bring together dealers, certifications and SEPA processing, getting clear up front who needed which screen and in what order the steps had to follow each other was no unnecessary luxury. It prevented a missing role or integration from surfacing halfway through the build. You can read more about that case on our projects page.
That same discipline, first getting clear on exactly what is needed and having it tested before the build starts, is not something an AI tool takes over just because it types faster than a developer. It is a step you organise yourself as a client or project lead, regardless of how the application is ultimately built.
What does this mean for your business?
A few concrete steps to combine AI speed with care, instead of letting them work against each other:
- Have an AI tool or a developer first create a clickable mockup with realistic sample data, in both phone and desktop formats, before any work starts on the real application.
- Write down beforehand what the goal is in one sentence, which data is never shared, and whether you will work with real or synthetic test data. That avoids discussions about scope afterwards.
- Have someone who did not build it review the mockup before moving on. Ask them to test a concrete outcome, not just to confirm that it looks good.
- Take stock of screens, roles and integrations before the build starts, and decide deliberately for each part whether it goes into the first version. That way you will not discover halfway through that something is missing.
The faster AI produces a working version, the more valuable it becomes to know first whether it is the right version. Would you like to know what that looks like in your situation, with or without AI support? Feel free to get in touch.