A Different Workflow
In the previous article, I explained why I don't believe one prompt will ever build serious software. AI can already generate applications, and it will only become better over time. But building software has never been just about writing code. It's about understanding a problem, making hundreds of small decisions, and improving a product over time.
So, if one prompt isn't the future, what is?
Well... I think the whole workflow will change. Not overnight, and probably not in exactly the way many people imagine, but I do believe we're moving towards a completely different way of building software.
Today, a typical project starts with meetings. People discuss ideas, collect requirements, prepare documentation, create mock-ups and eventually hand everything over to developers. Then developers write the code, QA tests it, feedback comes back, requirements change, and the whole cycle repeats again.
I don't think those stages will disappear. I think AI will become part of almost every one of them.
Requirements Will Become Much Better
Let's start with the first layer we talked about in the previous article. This is where ideas are created. Personally, I think this is one of the areas that will change the most.
Imagine describing a new feature to an AI assistant. Instead of simply writing down your ideas, the assistant starts asking questions you may not have considered. What happens if the user loses their internet connection? How should permissions work? What if two users edit the same data at the same time? Should the system support multiple languages in the future?
That's not generating code. That's improving the requirements before a single line of code is written.
And honestly, I think that's incredibly valuable. One of the biggest reasons projects become expensive isn't because developers write bad code. It's because the original requirements were incomplete. Every missing detail eventually becomes another discussion, another change request and another delay.
If AI can help us discover those missing pieces earlier, the entire project becomes more predictable. Developers receive clearer requirements, QA knows exactly what should happen, and clients spend less time changing things halfway through development.
Developers Won't Disappear
Now let's move to the second layer. This is where developers come in.
Will AI write much more code than it does today? I think so. In fact, we're already seeing it happen. Developers are using AI to generate functions, explain unfamiliar code, create unit tests, find bugs and even suggest better architectures.
But I don't think developers will simply press a button and watch AI build everything automatically.
Instead, I think their daily work will gradually change. Less time will be spent writing repetitive code. More time will be spent reviewing AI-generated solutions, improving them, connecting different systems together and making technical decisions that require experience.
In other words, developers may spend less time typing and more time thinking.
The Role of Senior Engineers
This is also why I don't believe experienced engineers will suddenly become less valuable. Actually, I think the opposite may happen.
AI is already very good at generating solutions. The difficult part is knowing whether those solutions are actually the right ones.
A senior engineer has usually seen the same problem many times before. They know why one architecture scales better than another. They know where security problems usually appear. They know which shortcut will save a week today but create six months of technical debt later.
I think that kind of experience will become even more important. AI can suggest ten different solutions in a few seconds, but somebody still has to choose the best one. Somebody has to understand the business, the users and the long-term consequences of every decision.
Maybe the job title won't change. Maybe we'll still call them software engineers. But I think their role will slowly move away from writing every line of code and towards reviewing, guiding and improving what AI produces.
Where I Think We're Heading
Sometimes people ask whether AI is like replacing the driver of a car.
I actually think it's closer to modern autopilot systems.
An autopilot can already do an impressive amount of work. It can keep the car in its lane, maintain speed and react to many situations much faster than a human. But when something unexpected happens, the driver still needs to understand what's going on and make the final decision.
For me, AI in software development looks very similar.
It will handle more and more of the routine work. It will become faster. It will become more accurate. And developers will probably trust it more every year. But I still think someone needs to understand the bigger picture, ask the right questions and recognise when the AI is confidently making the wrong decision.
Of course, I could be wrong. These are only my predictions based on what we're seeing today. Technology changes incredibly fast, and every year brings something new that surprises all of us.
But if I had to make one prediction, it would be this: I don't think AI is replacing software engineers. I think it's changing what software engineers spend their time doing. And those are two very different things.
In the next article, I'd like to look at something that worries many developers. If AI keeps getting better every year, will companies eventually need fewer engineers, or will they simply expect much more from the teams they already have?