The work was never just code

Today I complete 16 years working at AI/R. In this time I already saw the company reinvent itself at least five times, always following the changes in technology. I’ve been working with software development since 2003 and this made me stop and reflect about the cycles I already crossed, promises of transformation and real changes in the way we build software. This current moment, with AI everywhere, is one more of these cycles. Ok, it’s not just one more. It is by far, the most transformative and impactful of all.

It is almost impossible to talk about software development today without falling into the AI discussion. How many lines of code it generates, how much time it saves, how many people it “replaces”. This kind of debate is understandable, but it already starts looking to the wrong place.

After so many years working with software development, the feeling I have is that AI is not changing the essence of the work. It is showing, in a very open and clear way, what was always fragile in the way we build software.

The market always liked easy metrics. Lines of code, velocity, story points, commits. Remember “function points”? Yeah… AI just gave a new life to this behavior, now with nicer charts. But productivity is very different from quantity or volume. What really moves the needle is better product, less rework, more conscious technical decisions, teams less reactive and more questioning. This is still true, with or without AI.

What really changes is where human effort makes difference. Coding was never the hardest part. Hard is to decide what to code, why to code, and when not to code. Last week, in a bar talk with some friends, we said that today we can build almost anything and launch products in just one week. AI accelerates execution, but it does not replace perception and judgment about when and what to do. On the contrary, it increases the cost of bad decisions. Making mistakes fast became cheap, but fixing poorly thought decisions in corporate environments is still expensive.

I also see a lot of anxiety about reducing team size. But we already saw parts of this movie before. Whenever a technology makes something more efficient, the total usage tends to increase. With AI, the capacity to explore ideas, hypothesis and solutions grows a lot. This results in more product, more tests, more experiments, but also, many times, more organizational complexity. The bottleneck stops being code production and becomes alignment, architecture, long term vision and management.

In this scenario, the role of technical leadership changes a lot. It requires less control of details, less line by line review, and more definition of standards, limits and principles. The tech lead stops being the “best programmer in the room” and becomes the person who helps the team ask better questions, interpret answers and make coherent decisions in a bigger context. Yes, today more than ever, the tech lead needs to understand business and the impact that each implementation from the team brings to the company.

Architecture also becomes relevant again. Not as pretty diagrams, but as a decision tool. AI generates plausible solutions all the time, it makes everything work. But does this implementation really talk to the architecture? This is the filter that separates what is plausible from what is sustainable. Companies that don’t have this become hostage of solutions that work today and cost a lot tomorrow, now in a much bigger scale.

Another point that is not much talked about is maturity. AI does not level teams up automatically. It amplifies what already exists. Mature teams become faster and more consistent. Immature teams become faster to produce misalignment. The difference is not the tool, but the clarity of processes, quality criteria and technical responsibility.

The market will be less obsessed with “how much code was generated” and more focused on who can transform this capacity into real product. Companies that know how to use AI to reduce noise, increase clarity and strengthen decisions will stand out. The others will just deliver the same pains, only faster.

In the end, AI does not reduce the importance of people, it increases it. Because when execution becomes easy, thinking well becomes the most scarce asset.