Your own dashboard, a simple CRM or a first app prototype. With AI, you can now take an idea from concept to a working demo without even being a developer. What changes when you add real data and your colleagues or customers start using it? Tomáš Lodňan, CEO of GoodRequest, discussed the opportunities and limitations of vibe coding on the ‘Žijeme Sci-Fi’ podcast.
From idea to demo in no time
You have an idea for a tool that would make your work easier. Instead of spending ages describing the task, you open an AI tool, explain what you need, and gradually tweak the result. This is what ‘vibe coding’ looks like in practice: software is created using natural language and rapid experimentation.
For businesses, this is a useful way to test an idea. A clickable demo often reveals more than a few pages of a brief. You’ll discover what’s missing from the process, what you don’t actually need, and how people should use the tool.
In the interview, Tomáš describes his experiences at GoodRequest. Clients now arrive with their own demos, which they can use to show what they want to build. Thanks to the prototype, they can start with a more specific brief and develop it further into a fully-fledged product.
‘It works for me’ is the beginning, not the end
When testing, you’re mainly checking whether the application does what you’ve asked it to do. Once the team starts using it, further questions will arise.
Who has access to specific data? What happens when several people use the tool at the same time? Who will fix it if a change to one feature breaks another?
In the podcast, Tomáš highlights precisely this difference. Creating a functional app and designing reliable software are not the same task. Security, backups, access rights and long-term maintenance may not be immediately apparent. In real-world operation, however, you need them.
Just because an application displays data correctly doesn’t mean it protects it properly.
Three situations where it’s worth taking things slowly
You don’t have to stop experimenting. The important thing is to recognise when it’s no longer just an experiment.
- You’re entering real data. Customer details, payroll data and employees’ personal information require more care than test data.
- You’re making the application accessible via the internet. The login screen alone does not guarantee that data and access are securely configured.
- You’re integrating it with other systems. Access keys and permissions can also open the door to data outside the application itself.
Having another AI model carry out a check can be a useful first step. However, a professional security audit will examine the solution in greater depth. If your business already relies on the tool, you need to know whether it operates reliably and who is responsible for its operation.
How does a CEO use AI?
In the interview, Tomáš doesn’t just stick to giving recommendations. He also shares his own experiments: a financial dashboard and an AI agent that helps him with notes from one-to-one meetings and with feedback on his style of conducting interviews.
In his view, companies should actively test new tools, and management should lead by example. At the same time, it’s important to know when an experiment needs a second pair of eyes.
Listen to the full interview
Where does a useful prototype end and a product requiring professional attention begin? How is the work of developers changing? And what should you check once your company has its own AI-powered app?
Listen to the full episode of the ‘Žijeme Sci-Fi’ podcast with Tomáš Lodňan, CEO of GoodRequest.
