Raimo Seero, Chief Technology Officer at Uptime
Vibe-coded software solutions may look visually convincing, but someone without technical experience will not know how to spot the security risks hidden inside them. Companies should, however, encourage the creation of vibe-coded prototypes, because they make it possible to test ideas quickly and significantly speed up collaboration between developers and clients.
Where software development used to be firmly the domain of developers, vibe coding has now brought it within everyone’s reach. ChatGPT, Claude, Cursor, Bolt and other similar tools now make it possible to create applications simply through conversation.
The results are often visually convincing and, at first glance, fully functional. It is, therefore, no surprise that many companies have started to believe that software development simply means chatting with a chatbot. In reality, things are more complicated. Understanding that difference determines whether artificial intelligence genuinely helps a company or eventually turns into a headache.
The building will not withstand the storm
The main risk is that AI produces code without a deeply considered architecture or security layer, which means the solution cannot hold up under real-world load. It can be compared to building a roof for your house based on YouTube videos. The roof may even stay in place, but will the structure withstand rain, snow and wind? Unfortunately, that often only becomes clear when it is already too late.
The same applies to software. With vibe-coded solutions, security risks often materialise later – when customers, the company’s team or partners are already using the solution in practice. A person without a technical background will not notice the holes hidden in the code, because they are focused on the button visible on the screen, not on what is happening behind it.
When an idea emerges in a company for improving a process, the first step should be to create a prototype rather than hold another meeting.
For example, imagine that someone in a company vibe-codes a small internal website for sharing information with the team. Job application forms are added to it. Then someone asks for the HR department to be able to upload candidates’ CVs. At that point, the system is processing personal data: the solution must comply with GDPR, the data needs to be protected, and access rights must be managed. A person without a technical background may not think about this, and without the necessary input, artificial intelligence will not know how to handle it either.
A non-technical developer may not notice, or even know, when that line has been crossed. Understanding architecture, security requirements and system dependencies requires experience. Using artificial intelligence does not replace that experience.
Where AI really shines
This does not mean vibe coding has no value. It does – but its value lies elsewhere: in prototyping.
A prototype is not a finished product that is put into real use. It is a clickable model that shows how an application could work: in what order things happen and where decisions need to be made. It does not have to be beautiful or flawless, but it provides a first reference point for what the real solution might look like. People think visually, and when a specialist sees their idea on screen, it becomes much easier to notice missing data fields or logical errors that may have seemed reasonable on paper.
It also speeds up collaboration between the client and the development team. Instead of merely describing an idea, the client can show the developer what the final result could look like, and the developer can more quickly assess whether that result is actually feasible. In a way, it is an inexpensive insurance policy that helps avoid a situation where developers spend months building a perfect system, only for it to require immediate reworking the moment it meets reality.
Companies should encourage prototyping at every opportunity. When an idea emerges for improving a process, the first step should be to create a prototype rather than hold another meeting. The best results come when the prototype is created by the specialists who will actually use the new solution in their daily work, because that is how you can make sure the tool will ultimately be useful.
Artificial intelligence has democratised software development, and that is undoubtedly good news. It significantly accelerates collaboration between the business side and the IT department. But companies need to draw a clear line between a visual sketch and production-ready software. AI may be able to build an application, but in the end, the responsibility lies with the company that puts it into use.