AI: new opportunities for better software?

Last updated April 11, 2026

In the software industry, we have become accustomed to accepting quality issues as the norm: we have too few automated tests. Changes are often riskier than they need to be. Refactorings are put off until later because we don’t make the time for them. As a result, codebases become harder to modify with every iteration.

We keep hearing the same excuses over and over again: we don’t have time, we’re under too much pressure, and besides, our application is so terribly complex. Whilst all of that is understandable, it’s still a problem.

Quality issues are systemic

Poor code quality rarely arises because people are incapable of building good software. It arises because the reality of our working lives regularly penalises good technical decisions. Those under intense delivery pressure do not prioritise investment in test coverage, clean code structures or systematic code refactoring, because the system often creates different incentives.

That is why better tooling is so important. When tools help us to generate tests, prepare for safe refactorings, validate changes more quickly and identify vulnerabilities at an earlier stage, something fundamental shifts, and quality is no longer merely a matter of individual discipline. That is the real progress.

Speed is not the key factor

We often talk about speed when it comes to automation. I think that’s too short-sighted. The more interesting factor isn’t that we can generate code faster. The more interesting factor is that we have the opportunity to systematically produce better code.

And this could mark the start of a development that extends far beyond individual teams.

AI could make the industry more adaptable

If we train AI through iterative cycles featuring increasingly better code, improved testing, cleaner refactoring and more transparent architectural decisions, this could take our industry’s adaptability to a whole new level.

So far, software development has been knowledge-intensive, but only learns to a limited extent. Good solutions emerge in many places. But they spread slowly, haphazardly and often by chance. Much of the experience remains localised. Much of the excellence remains tied to individual people or teams. And too many mistakes are repeated time and again.

With better tools, we can break this pattern.

When best practices are not merely documented but actively integrated into day-to-day work through tools, quality scales differently. It then depends less on heroics, less on individual senior engineers holding onto tacit knowledge. And, above all, quality no longer arises by chance or luck.

Stattdessen könnte eine Umgebung entstehen, in der gute Lösungen wahrscheinlicher werden, weil sie technisch unterstützt, im Prozess verankert und durch lernende Systeme verstärkt werden. Darin liegt für mich die eigentliche Hoffnung.

AI not only amplifies quality, but also weaknesses

However, if we miss the opportunity for continuous improvement, the opposite could happen: instead of entering a spiral of improvement, we would find ourselves in a spiral of decline. This is because AI does not automatically enhance quality; rather, it primarily amplifies the patterns, decisions and routines that we feed into it. If we reproduce poor code, weak tests, messy structures and short-term technical compromises on a large scale, we are not only speeding up development but also accelerating deterioration. Technical debt would then systematically increase.

I don’t believe that AI will replace us. Instead, I believe there is a real opportunity for AI to help us reach a higher professional standard. That it automates routines, reduces risks and makes sound technical decisions more cost-effective. That it turns quality standards into concrete practice.

To achieve this, we must use our tools in such a way that our industry systematically learns to write better code from good code. If AI shortens the learning cycle from one generation of developers to a model’s training cycle, we gain more than just productivity. We gain the opportunity for a software industry that consistently learns from its best solutions, spreads quality across the board, and no longer leaves technical excellence to chance.

That would not only be more efficient, but above all more mature.