Will AI Replace Mid-Level Software Engineers?

No, I don’t think we’re there yet. However, mid-level software engineers should take this as a warning and use it as motivation to learn and adapt. Check out my advice.

Feb 02, 2025

Featured on Software Testing Weekly #257

I came across this Reddit post: “Mark Zuckerberg believes in 2025, Meta will probably have a mid-level engineer AI that can write code, and over time it will replace people engineers”, links to this video.

“Probably in 2025, we are Meta as well as the other companies that are basically working on this, are going to have an AI that can effectively be a sort of mid-level engineer that you have in your company that can write code. “ - Mark Zuckerburg.

I think Mark Zuckerberg is exaggerating/overly optimistic about this. Here, I offer some thoughts and alternative perspectives.

1. Why Mid-Level? Not Senior or Junior?

Mark’s comment suggests that mid-level software engineers are less creative and often involved in repetitive or mechanical tasks, which holds some truth.

Here’s my interpretation of Mark Zuckerberg’s underlying message (please note, I don’t agree with his predictive outcome):

Junior engineers bring energy and fresh ideas. Many cutting-edge companies, like DeepSeek and SpaceX, are heavily staffed with engineers who are less than two years out of university. Additionally, their salaries are relatively low.
Senior engineers—especially top-tier ones—can be exceptionally valuable, sometimes performing at 10X or even 100X levels. AI will never replace them, and they play a crucial role in mentoring junior engineers.

Don't forget that Mark Zuckerberg is an entrepreneur who employs over 70,000 people, many of them are mid-level software engineers earning well—around $300K annually. Naturally, if a viable replacement existed, he would likely pursue it.

Frankly, as a software startup owner, I’m not fond of hiring mid-level software engineers, either. If I could attract a world-class senior engineer—someone capable of developing and maintaining a large E2E test suite (which is extremely rare)—that would be fantastic.

“In my experience, great developers do not always make great testers, but great testers (who also have strong design skills) can make great developers. It’s a mindset and a passion. … They are gold”.
- Patrick Copeland, Google Senior Engineering Director, in an interview (2010)
“Testing is harder than developing. If you want to have good testing you need to put your best people in testing.”
- Gerald Weinberg, software legend, in a podcast (2018)

However, knowing that's nearly impossible, I much prefer hiring new motivated graduates or juniors whom I can mentor and instill the right mindset in. I have done that, and the outcome was great.

I have already published an article on this: Practical Advice for Hiring a Junior Test Automation Engineer,

2. Why do I think “AI Replacing Mid-Level Software Engineers is unlikely in 2025”?

In terms of producing code, AI might be close to on par or even excel on generic tasks, such as:

Create a TODO List app in React
Create a Ruby on Rails app with built-in user authentication
…

However, we know that coding is just a small part of software development, and AI still falls short in several key areas:

Interpreting Requirements – Customers and business analysts often struggle to articulate their needs clearly. Understanding and refining vague requirements is a critical human skill.
Integration – Unless AI handles an entire project development, it must integrate with work done by human developers. This process is complex and cannot be underestimated.
Handling Changes – Agile embraces constant and frequent changes, something AI struggles to manage effectively.
Maintenance – Post-deployment changes require special care. A single mistake can corrupt critical data beyond recovery.

I know some AI enthusiasts might want to debate this, so let me illustrate with an example. I’m both an accomplished programmer (winner of an international programming award) and an expert in E2E test automation (a hands-on coach and author of 13 books in the field). I first heard about ChatGPT in December 2022 from one software tester, who claimed it could revolutionize software testing and automation. I did some research. I disagreed and went on to write several articles on the topic.

Check out this article, ChatGPT is out for over 6 months. Have you implemented real test automation with AI? (published in 2023)

The latest version of ChatGPT, now over two years old with a few releases, is probably 100+ times more powerful than when it was first released. To put it simply, if someone claimed AI test automation was so useful, let’s say, even merely 5% useful in late 2022, in theory, it should be close to 100% now. However, when it comes to the projects these AI test automation enthusiasts are working on, I’d bet their E2E regression testing is still done manually and is still expensive, error-prone, and not frequent (a real agile team does full E2E regression testing several times a day, if pass, deployed to production).

Advice to Mid-Level Software Engineers

Regardless of AI advancements, mid-level software engineers should continue to refine their skills in software design, debugging, and testing. By becoming a truly impactful senior engineer—beyond just holding the title—you can ensure that AI won’t easily replace you.

How can a mid-level developer gauge their progress toward senior technical capability? Assess yourself based on the following:

Flexible Design – Your designs can easily adapt to future changes requested by customers.
High Productivity – You consistently improve your efficiency, achieving 2X, 5X, or even 10X gains compared to your past performance.
Proficiency – You are skilled in using essential tools like coding IDEs, debuggers, test automation frameworks, and continuous testing servers.
Automation Mindset – You embrace automation to streamline workflows and reduce manual effort.

I made this transformation (to 10X) after embracing E2E (UI) test automation. Here is my Advice #1: Learn E2E (UI) Test Automation, seriously. If you do, you'll thank me later. However, I know, most won’t take any action.

Check out my article, “Tim Ferriss, the world’s most famous life coach, advises learning coding to automate!”.

From my observation, many mid-level software engineers in recent generations lack motivation for structured learning, such as reading books and doing exercises. Instead, they often rely on copying code from Stack Overflow—or now, AI tools—without deeply understanding it. Naturally, these software engineers would be the first to be replaced.

Advice #2: Start your Side-Hustle project(s)

Your employer will constantly look for opportunities to replace you with AI, automation tools, or more cost-effective alternatives, such as outsourcing to countries like India. And yes, sometimes bosses make silly decisions—unfortunately, you could be the casualty.

Now, consider a different perspective: What if you leveraged AI and automation to build your own apps? Start as a side hustle in your free time. One day, you might become a solopreneur or startup founder—essentially replacing the boss with yourself.

If you are the kind of person don’t like to work for others, you find this newsletter helpful.

Related reading:

The Agile Way

Discussion about this post