I’ve tried to ignore it, but the AI hype has grown louder and louder. From my beginnings in the tech industry over twenty years ago, there was always some revolutionary new thing just around the corner movers and shakers in the industry couldn’t stop talking about.
First it was the dot-com hype, then Web 2.0/ social networks, cloud technology, Web 3.0, crypto-currency, AR (Augmented Reality), VR (Virtual Reality) and now finally AI. Going by the latest reports, AI will do everything from your laundry to driving you to work.
Except in my case I wont be driven to work – because I will have been replaced by an intelligent AI, that can spit out code, deploy and mantain it in perfect alignment with business requirements. As a business owner, or C level manager the benefits are hard to ignore. No more pesky well paid software engineers to reduce the margins. Massive profits from exponential growth of sales from a product built through automation. Increased returns for shareholders, they hit KPIs and unlock bonuses. Trebles all around.
I apologise. I lied.
Adwin Anderson
So going back to the article title. I know I said “AI Will Replace You”. I apologise. I lied. AI won’t replace you as a Software Engineer. It might make you temporarily unemployed, but don’t worry your career wont be over.
The truth is, there isnt a shortage of software engineers. Millions of them appear every year. If you can open an IDE and build a basic CRUD system – you have a decent shot at getting some kind of tech job. What there is a shortage of, is good software engineers. And you only become a good software engineer by either having an extremely strong grasp of the fundamentals, and some of the softskills, or building up battle tested experience working on many different types of projects with different teams of people across multiple companies. The best software engineers have gone down both of these paths – and guess what it takes years.
The main reason why I can’t see AI taking over is because software is written for human benefit, not machines. The output needs to make sense and be usable for some type of human activity. This means whatever the AI creates needs to be checked, tested and reviewed by humans, from engineers, to testers and product managers. Sure, a software engineer in 2025 might not need to spend as much time writing syntax in an IDE, but they’ll damn sure have to read every line, run every test and understand what the AI is throwing out.
I hate to burst the bubble, but the coding life isn’t for everyone. Don’t listen to the hype being pushed by your favourite dev influencer on your go-to social app. You can’t just do a bootcamp for a few months then get a job where you get six figures a year, a four thousand pound laptop, a smart phone of your choice, then decide which country you’ll be working from next week. (For a brief period in 2020-2021 this was actually possible). You are either being lied to, or sold to. Or both. I probably spend less than 20% of my time writing code. Most of the time im either fixing bugs, dealing with compliance issues, maintaining servers or doing a myriad of other tasks that come with maintaining a complex multi-tenanted application.
AI isn’t the silver bullet to all our problems. Software engineers are still a critical part of the software development process and its difficult for me to see that changing anytime soon.
It’s better to see Artificial Intelligence as Augmented Intelligence. Generative AI, and AI agents can only improve the productivity of a software engineer, not replace it. Software has become a critical part of modern society and runs many critical industries that underpin the basic functions of modern society such as transportation, banking, defence, utilities and more. All of these are too important to leave in the hands of an opaque blackbox algorithm no-one quite understands.
Planes, trains and to an increasing degree road vehicles can operate autonomously most of the time, but they still need a human in control for that small percentage of situations as any mistake can lead to severe, and often fatal consequences. Deleted code can be restored, rebuilt and refactored. Deleted humans – not so much. Its the same with software. As it becomes increasingly critical boring old school things like regulations and insurance requirements have forced companies to grudgingly keep armies of engineers in work. AI might be able to do most of the boilerplate work, scaffold an app, write some basic tests but it can’t yet build an end to end deployment pipeline, integrate third party APIs, and build secure network architectures. And it definitely can’t do this in a way that fits a business commercially proprietary logic requirements. For that, well paid humans will still be needed.
The question is how many?
