Will AI Replace Software Developers?

Photo by Florian Olivo on Unsplash

First things first: this is just my humble and personal opinion. I’m not an expert in generative AI or economics, but I find this debate incredibly fascinating — and honestly, it’s fun to make predictions and look back on them in a few years 🙂

Right now, as always, the conversation seems highly polarized, some statements that I have read lately:

1. “Thanks to AI, engineers will get rid of the boring parts and focus on creative work.”
2. “It’s impossible to replace engineers because AI just predicts — it doesn’t truly build.”
3. “In two years, 90% of engineers won’t be necessary.”
4. “AI will give rise to new business models we can’t even imagine yet.”

With all due respect, I believe most of these positions are exaggerated and fragile, although all of them have a part of truth. As with many things, the truth might lie somewhere in the boring middle. Here’s my take:

1. The argument that AI will free us from boring tasks so we can focus on creativity is, in my view, a bit naive. If AI makes an engineer 50% more productive, then companies will likely need 50% fewer engineers to deliver the same business value. I’m sorry to say it, but I doubt most companies will pay us just to have more fun at work. The key question is whether investment in digitalization will remain stable — if so, we might just be able to do more with the same resources. Maybe it’s my bias, but I still see many untapped opportunities to create value through technology. I will go deeper about this point later.
2. Dismissing AI as merely a predictive system doesn’t feel accurate either. That’s an oversimplification. After all, isn’t our own brain essentially predicting the best lines of code to write, based on our knowledge? The boundary between creative reasoning and just “spitting out tokens” is thin.
3. “If 90% of code is AI-generated, then 90% of developers won’t be needed” — that’s another fallacy. Just because most code is assisted by AI doesn’t mean it won’t need a human behind it. Taken to the extreme: 100% of code is written using keyboards… and yet here we are. That said, I believe today’s AI paradigm is inherently limited when it comes to fully replacing humans end-to-end. The lack of determinism (for now) prevents us from scaling full automation. Of course, that doesn’t mean efficiency gains won’t be significant. We’ll see what happens with new paradigms (buzzword alert: agents) — it seems like only a matter of time.
4. “AI will bring about new business models we haven’t imagined yet” — perhaps, but this feels like wishful thinking. It’s true that this has always happened with previous technological shifts, but there are also economic models that challenge the idea of infinite growth. While past innovations have often led to economic expansion and job creation, it’s unclear where that limit might be. Personally, I see saturation even in entertainment industry.

Obviously, all these narratives are shaped by vested interests — especially the overly optimistic ones, which often come from companies that directly benefit from the hype surrounding AI’s potential. It goes without saying: we should approach those messages with caution.

I also have somewhat contradictory feelings myself. On one hand, I recently tried some AI coding tools and was blown away — this is clearly going to change the way we build software, forever. On the other hand, I can’t shake the feeling that many of us in the industry know too well: 90% of the time goes into the last 10% of the work. It looks like we’re almost there with having AI build software using just natural language prompts… but maybe that last stretch is what will take the longest. Think of autonomous cars — they’ve been “just around the corner” for the past five years.

As for the future of jobs in software development — I don’t think anyone really knows. With my limited background in economics, I believe macro-level trends (consumer demand, international conflict, pandemics, geopolitical instability…) are essentially unpredictable and largely beyond the control of 98% of the population.

So, going back to point (1): yes, we’ll likely need fewer engineers to build the same capabilities. But will overall investment in tech stay the same? If so, maybe we’ll just invest in more ideas with the same number of people. There are many potential scenarios:

• There’s the Jevons Paradox, which everyone seems to be referencing lately. It says that when something becomes more efficient (in this case, software development), usage actually increases instead of decreasing.
• Until recently, digitalization often meant adopting SaaS. But if building becomes easier, companies might find it more worthwhile to develop their own custom capabilities.
• …

In summary:

Yes, I do believe AI will eventually replace some software developers — just like any technological evolution has replaced certain types of jobs. The real question is when.

In the short to medium term, I think the impact will be gradual and far from dramatic. As with many big changes, it will creep in little by little, almost without us noticing. Like I mentioned earlier: the last 10% takes 90% of the time — especially in a deterministic domain. So, let’s stay calm and not give in to panic or anxiety. There are a lot of incentives out there pushing us toward FOMO and urgency.

In the medium to long term, it will depend on broader, more controversial questions — like whether capitalism can sustain continuous growth, or whether we’ll reach such high levels of efficiency and automation that we’ll need to rethink everything (universal basic income, drastically shorter workweeks, etc.). If that happens, any cognitive job could become replaceable.

But what’s certain is that AI will continue to improve exponentially and will definitely transform software development roles. We must embrace that reality — not ignore it — but we can do so calmly and thoughtfully.

And the million-dollar question: Would I recommend my kids study software development if they showed interest?

Absolutely yes. Building things is a joyful experience. Even if we don’t know what building software will look like in the future, the process itself is deeply rewarding. And I personally believe that to get the most out of abstractions, it’s essential to understand the foundations.

Super interesting times ahead, let’s have fun coding for now!

Thanks for reading — especially if you made it through all my rambling. Opinions are my own 😊