Three Reasons Why AI Isn’t Replacing Software Engineers

The impact of AI on software engineers is impossible to deny. Perhaps more than any other industry, software has been entirely transformed by this new technology and the future of software will never be the same.

But will AI replace software engineers entirely? We don’t think so and today we’d like to break down three reasons why.

1. AI Can Code But it Can’t Design

Let’s be clear about one thing right out the gate. AI is good at coding. In fact it’s better at coding than it is at just about anything else.

The reason for this is that AI tends to produce very predictable text, following common patterns in language use. This is a problem for writing but it’s exactly what you want from a piece of code.

Good code is predictable. Naming conventions should always match. The structure should be standardised. Anything not immediately intuitive should be carefully marked out with comments. All of which AI is naturally good at.

But there’s one major problem with AI code. AI lacks context.

For example, software engineers spend a lot of time thinking about user experience and interface (UX/UI). This part of the design process is all about giving the user an intuitive experience. There are many principles that go into UX/UI but at its core, you need to be familiar with what it’s like to use software. Put simply you need to know where to put which buttons so that users will find the tools they’re looking for.

AI doesn’t have that. It’s never been a person. There are just too many micro-interactions involved for an AI model to design well. At best it can copy and learn from other designs but that can only get you so far. Good UI/UX is very tailored to the application in question.

And it’s not just the user-side either. AI often lacks context about how/why you might want to extend your software. It will make assumptions about the way modules fit together that aren’t always true. It’s excellent at a lot of the micro-stuff but, left to its own devices, AI just isn’t very good at assembling the macro.

2. Someone Needs to Overview AI Code

And it’s not just about the design side of things either. Even the tasks AI is good at like writing boilerplate code and coming up with naming conventions, ultimately require some human oversight.

If you look at the impact AI has had on software engineering jobs up until now, it should come as no surprise that it’s the juniors who are getting replaced while the seniors are expected to do more work in fewer hours. This is because there still needs to be someone to approve and sign their name to the work, even if that work was done/assisted by an AI.

For this to change there would need to be a cultural shift as well as a technological one. As it stands, people don’t fully trust AI work. AI is known to make mistakes and, unlike a human, it’s not going to spot and correct its mistakes. Where a human programmer might take a step back and question their own decisions, an AI model will just plough ahead, bolting on solutions and building an incredibly messy codebase.

So while the future of software engineering jobs might be a smaller market, there will always be demand for people who know how to program and can oversee an application through development.

3. There’s Always More to Do

It’s worth noting that no application is ever truly complete. Making software is a constant balance of building new features while maintaining and bug fixing old ones.

On the one hand, the more users you have, the more likely it is that you’ll encounter an edge case bug. On the other, even if your app is entirely bug free – which is very unlikely to happen – it won’t stay that way forever. Software is always built on other software, on packages and operating systems and a vast network of code that you’ll never touch.

All of that software is itself constantly being updated. New Windows update? Get ready to fix some bugs? Need to update to a new python library? More bugs incoming.

So with all that in mind, someone will always have the job of working out what’s gone wrong and how to fix it. Even if AI sped up the work tenfold, developers would never run out of tickets.