The Question Everyone Is Asking
In the previous article, I talked about how I think the software development workflow will change. AI will become part of almost every stage of development, from writing requirements to generating code and reviewing solutions. But that naturally leads to another question.
If AI keeps getting better every year, what is left for engineers?
Well... I think that's actually the wrong question. The better question is this: what can AI do really well today, and where does it still have limitations? I don't think we can make good predictions about the future without first understanding both sides.
AI Works with Probability
One thing that's easy to forget is how these models actually work. They don't think like humans. They don't understand the world the same way we do. In very simple terms, they're making predictions based on probability. Given everything they've learned, they try to generate the answer that is most likely to be correct.
Most of the time, that's exactly what makes them so useful. They can explain complicated topics, generate code, summarise documents and answer questions in just a few seconds. Honestly, I use AI almost every day, and I think it's one of the most useful tools we've had in software development.
But probability also has a downside.
Sometimes AI sounds completely confident while being completely wrong. And when two possible answers are very close, maybe one has a 51% chance and the other 49%, it's surprisingly easy for the model to choose the wrong one. That's not because it's "stupid." It's simply the way these systems work today.
That's why I don't think we should treat AI as an unquestionable source of truth. We should treat it as an incredibly knowledgeable assistant that still needs somebody to review the final result.
Experience Is Different from Knowledge
Now... here's something I find much more interesting.
We often talk about knowledge as if it's the only thing that matters. But I don't think that's true. In many professions, experience is just as important, and sometimes it's even more valuable.
Imagine an engineer who's been designing suspension systems for twenty or thirty years. They may look at a design and immediately say, "I don't like this material. It will probably become a problem after a few years." Maybe that exact information isn't written anywhere. Maybe it's something they've learned after seeing hundreds of real products over the course of their career.
That's difficult to replace with documentation alone.
The same thing happens in software development. A senior engineer often recognises a bad design within a few minutes. Not because they've memorised another programming book, but because they've already seen a similar solution fail several times before.
I think that's where experience becomes incredibly valuable. AI can suggest many solutions, but somebody still has to recognise which one is likely to work in the real world.
AI Doesn't Experience the Real World
Let's take this idea a little further.
A carpenter doesn't just know how to work with wood. They can see the grain. They can feel the material. They know when something doesn't look right. A mechanic can often hear that an engine isn't working properly before any diagnostic tool reports an error.
Those decisions come from years of experience combined with real-world observation.
Today's AI doesn't experience the world like that. It doesn't see materials, hear unusual sounds or develop intuition through physical experience. It learns from data, which is incredibly powerful, but it's still different from interacting with the real world every day.
Maybe that will change one day. Robots are becoming more capable every year, and future AI systems may collect their own experience through sensors and continuous learning. But I think we're still a long way from that point.
My Prediction
So... where does that leave us?
Personally, I don't think the future belongs to either humans or AI working alone. I think the strongest combination will be people who know how to use AI effectively while adding something AI still struggles to provide: experience, judgement and context.
Knowledge is becoming easier to access every year. Experience isn't.
That's why I don't believe experienced engineers suddenly become less valuable. If anything, I think their role becomes even more important. They'll spend less time writing repetitive code and more time reviewing ideas, making decisions and guiding AI towards better solutions.
Of course, these are only my predictions. I could be wrong, and technology has surprised us many times before. But based on what we can see today, this feels like the direction we're moving in.
In the next article, I'd like to look at a few real examples that changed my own thinking. We'll talk about Stack Overflow, one of the biggest knowledge-sharing platforms in our industry, and why I think the rise of AI completely changed its role. We'll also look at the Ford story that many people discussed and what I think we can actually learn from it.