Why It’s Too Late to Learn Automation
The idea for this post came to me during a regular meeting with my fellow mentors, SDETs from several international companies. We were discussing the future of the QA Automation market and reached some rather interesting conclusions.
But before I share our findings, let’s wind the clock back five years and remember why almost every manual tester dreamed of breaking into automation:
- More engaging and complex tasks: Essentially a hybrid of testing and development.
- Market demand: The vacancy ratio was roughly 70/30 in favor of automation engineers.
- Higher compensation: Automation required a broader skill set and continuous learning, which naturally commanded higher salaries.
For entry-level positions, it was enough to write basic code in Python, Java, or JS, possess structured thinking, and understand CSS/XPath, the DOM tree, Git, and base libraries like Selenium. This is why the path from manual QA to automated testing was so popular and didn’t take an eternity. I followed this path myself many years ago, and I can’t say it was insurmountably difficult.
In today’s market, automation engineers are still needed, but the role now usually implies a specialist who knows how to work with AI development tools—meaning, effectively, a Senior who tasks the AI and then ensures the resulting code is maintainable, clean, elegant, and fast.
Now, let’s look 2–3 years ahead. Obviously, AI models and tools (like Claude Code) will continue to evolve, eventually reaching the level of a generalist Senior Developer who can write code and cover it with tests depending on the task.
Consequently, the market will no longer need “pure” programmers, let alone traditional automation engineers.
So, what should we do? Should we quit IT immediately and “learn to weld” or become truck drivers, as some US tech chats suggest?
My colleagues and I believe there is a way out. Yes, AI models are already generating decent low-level code. However, looking a few years ahead, the issues we see now—the inability to create a cohesive, maintainable architecture and ensure the necessary level of quality—will likely persist.
After all, the quality of an AI model’s performance depends heavily on the volume of training data. While there are no issues with examples of working code for training datasets, there are far fewer descriptions of software architecture—and, more importantly, high-quality architecture.
So, someone will still have to oversee whether the software being developed is maintainable, scalable, and therefore cost-effective. In other words, the issue of quality will be elevated to a higher level.
In summary, your goal for the next couple of years should be to achieve a high level of expertise in system design and architecture, the ability to formulate both business and system requirements, and, of course, a deep understanding of how AI models work.
But before reaching that goal, you’ll need to level up your programming—not to write assertions for autotests, but to gain the ability to transition to that next architectural level.
The Plan:
- Level up programming in any language to a Senior level.
- Deep dive into architecture and system design.
- Study business analysis and a specific subject area.
Yes, balancing this with a full-time job and family responsibilities can be tough. But then again, who said staying in IT would be easy?
