"AI is going to replace me." You hear that fear everywhere. Yet AI doesn't replace: it displaces, transforms, repositions.
The risk of losing your job to AI comes less from AI itself than from our inability to question ourselves and identify our new role. What we need to produce doesn't change, it's the way we do it that changes. I learned this throughout my career and it took me years, because I was blazing a trail with no map. You inherit the path: a few months should be enough.
A tech enthusiast from a young age, I still remember the first wave of local AIML chatbots (the ancestor of the LLM) and the earliest publications I read on the beginnings of deep learning.
Still searching for my path at the time, I was already playing apprentice developer through HTML and PHP without even knowing it was a real profession. It was only two years later that the dev world would reveal itself to me with all the challenges it promised.
Starting my career in Japan at a small company as a junior dev, I kept in mind the story my teachers had repeated about the lumberjack who had to keep his axe sharp. I organized my tech watch as much out of curiosity as professional interest. I was thrilled by new PHP frameworks and the micro-revolutions in JavaScript, and what they promised. Experimenting and proposing new tools and approaches with each new project, I found myself promoted to tech lead, then team lead, in charge of training new hires and designing incoming projects.
My work having shifted from writing code to reviewing code written by others, I didn't have time to dig into the areas where I struggled, and the exciting part of my job was being assigned to others. I was dragging my feet on the way to work. But that's when the one who would become one of my most loyal collaborators arrived: GPT-3. With GPT, syntax review and bug hunting became almost instant, freeing me from one of my biggest workloads so I could get my hands back in the engine.
With GPT and Codeium, the only equivalent available at the time in my dev tool, I could steer code with just a few words in a comment. I could rewrite in an afternoon what had taken me several days, but cleaner, shorter, and more reliable. I watched my delivery timelines cut in half, feeling like I was taking off. GPT had also become my mentor. Able to find the information I was looking for instantly across pages of documentation, it also let me synthesize, rephrase, re-explain, all with infinite patience. In a few months, I had progressed more than in the previous five years.
My tedious workload lightened with each new version. The numbers spoke for themselves: it had taken 14 months to ship a product with 5 people. Version two, which required a complete rewrite, more ambitious and comprehensive, took only 7 months and 3 developers. Immediate sale and success. The advantage of AI was indisputable, overwhelming. But at the end of this period of disbelief, the real revelation came: AI was already better than me. Guided, sure, not yet autonomous, but that limitation could only be temporary. I had to face the facts: I was barely writing code anymore, I was supervising, directing. Code review was assisted, coding was assisted, I had become my assistant's assistant.
I then entered my first paternity leave. With the joy and demands of my first child, it became hard to find time to code. I would throw out ideas and GPT, then in its version 4, would execute. I was no longer guiding code, but a project. I had become, despite myself, a project manager. After months of looking for a new dev role (and a solid phase of denial), I decided to become a PM. Goodbye IT, hello PM.
This job scared me. Having to create specs, prototypes, write endless manuals... Upon starting, I discovered that all of this, AI could do far better than me. What was left for me to do? The heart of the job: understanding the needs, weaving a concept, sketching a solution. I was no longer a writing machine, I had (re)become a thinking machine.
Today, in 2026, I've given up the fight against AI. It's better than me at coding. That's a fact. And this is where people usually make the mistake of stopping. Is that all that defines me? Being a simple coding tool? What if I was only that tool because THE bottleneck slowing down every project is precisely writing code? Just because a task represents 90% of your work doesn't mean that task defines your work. Remove coding, and everything else remains: identifying the real problem, choosing the right approach and technologies, designing the architecture, making a myriad of micro-decisions, imagining improvements and new features the client thought simply impossible...
In the end, this is where my real value lies. Where my expertise shines. Today, I'm a fulfilled PM, thriving in my work, and I love the design challenges it offers. All the unimportant, time-consuming tasks, my agents handle those; I decide. And it's several years of denial, of inner struggle that I'm sharing with you, hoping to save you the price of the ticket on this journey. Fighting is pointless, because AI, whether we like it or not, is here, and has already begun the change. Instead of trying to stop progress, try to (re)discover the value you bring to your company. Together, let's lighten everything that doesn't require your expertise and refocus on what truly matters.