Are you thinking about creating a project from scratch using Artificial Intelligence?
It could be a business, an internal tool, a website, a digital service, or simply that idea you've had in mind for a while and haven't been able to build until now because you needed a technical team, a significant budget, or knowledge you didn't have.
The good news is that AI has lowered many of those barriers.
The not-so-good news is that being able to build almost anything doesn't mean that everything you build is going to make sense.
For me, that is one of the most interesting questions of the current moment. Artificial Intelligence provides us with an enormous execution capacity, but the value of the project still depends on our ability to detect a problem, design a solution, and get someone to want to use it.
AI has changed the speed, not the fundamentals
For years, turning an idea into something tangible required preparing a proposal, securing a budget, assembling a team, and waiting weeks or months to have a first version.
That process caused many ideas to die before we could check if they were any good.
Today we can prepare a website, a prototype, or an automation in hours. We can even modify it almost live after showing it to a user.
What used to require several meetings can now be resolved by building a first minimum viable product and observing what happens.
But the underlying process remains the same:
- Detect a need.
- Understand who it affects.
- Design a value proposition.
- Build a first solution.
- Test it with real people.
- Learn and iterate.
AI accelerates each of these phases, but it doesn't eliminate any of them.
Don't start by asking which tool you should use
When someone wants to start an AI project, one of the first questions is usually whether they should use Claude, Codex, ChatGPT, Gemini, or any other tool.
It's an understandable question, but it's not the first one we should answer.
First, we need to know what we want to build, for whom, and why it would have value.
The best tool is not necessarily the most powerful one. It's the one that allows you to move forward at the point where you are without adding complexity that you cannot yet control.
You can start with a simple tool to create a basic website, learn how a project is organized, and understand what pieces it needs. When that way of working feels too small for you, it will make sense to move on to more agentic systems capable of acting on numerous files and processes.
To put it into a simple example, learn to drive with a simple car first (Gemini, ChatGPT...) and when you feel safe, comfortable, and confident, you will be ready to drive an Aston Martin (Codex, Claude...)
What does vibe coding allow you to do?
Vibe coding allows you to develop software by relying on AI even if you don't have a traditional technical background.
This opens up huge possibilities. A person can create a website, automate a process, develop a recommender system, or prepare a tool for their team without waiting for an engineering department.
But AI does not eliminate the need to understand what you are building.
If you want to publish a website, you need to discover what a domain, a server, or a hosting system is. If you use user data, you must think about security and privacy. If you expect thousands of visits, you will need to understand how the infrastructure responds.
AI can explain every concept to you and help you execute it. However, you need to have enough judgment to ask the right questions and recognize when the answer isn't adequate.
Validate your ideas before building too much
Current speed can lead us to produce a lot before checking if anyone needs it.
Precisely because building is easier, we should validate sooner.
Prepare the smallest possible version that allows you to understand the idea. Show it to people who have the problem. Observe what they understand, what they use, and what they ignore. Change the solution based on what you learn.
It no longer makes sense to spend weeks debating a screen that you can build in an afternoon.
Many times it takes me less time to do it than to explain it. When a first version exists, the conversation stops revolving around abstract opinions and starts relying on something we can test.
A demo made with AI is not yet a product
Making something work once demonstrates a possibility. Turning it into a product requires much more.
The solution must be desirable for its users, viable as a business, and feasible from a technical standpoint. It must also be able to operate with reasonable costs, protect information, handle errors, and adapt if usage increases.
You don't need to solve the problems of a platform with millions of users from day one. We can start with a simple infrastructure and evolve it when there is a real need.
The important thing is to move forward step by step and maintain the capacity to adapt.
What is the most common mistake in AI projects?
The most common mistake when creating an AI project is thinking that the entire solution must revolve around it.
AI is not the value proposition in itself that allows us, automatically, to build what is in our heads.
The center of everything must be the problem we solve and how we solve it to provide value. Once we have this clear, we can decide where Artificial Intelligence can help us turn that idea into reality, and it is in this sense that Artificial Intelligence can allow us to research, program, automate, and generate content much faster and more efficiently.
Nor would I completely delegate to AI judgment, strategy, the empathy needed when detecting where value can be added, or the creative direction of our idea. Artificial Intelligence can offer alternatives, analyze information, and help us explore, but it is still up to us to decide what is worth building and why it might matter to someone and impact a business's bottom line.
What is the best tool for an AI project?
Artificial Intelligence can become a programmer, researcher, the best assistant and/or consultant in the world, or the best software engineer you have ever met. Because yes, Artificial Intelligence can vastly expand your capabilities and allow you to do things that until recently seemed out of your reach.
But the most important tool remains your knowledge, your creativity, your curiosity, and your judgment.
You don't need to start with the definitive project. Turning an idea into a business or a successful project usually depends on starting small and building something you can show, learning from user reactions, business impact, and the technological challenges you face—and that, undoubtedly, depends more on you than on AI.
Because yes, Artificial Intelligence has shortened the distance between having an idea and seeing it work, but navigating that path successfully still depends on you.
