For years, the software industry operated under one core assumption: writing code was the primary bottleneck in software development. As a result, companies poured enormous resources into hiring developers, refining development processes, and building tools that could shorten the journey from idea to product.
Then generative AI arrived.
For the first time, the industry gained a technology capable of solving the very problem it had spent decades trying to overcome. The ability to write code – once the industry’s most constrained resource – is now faster, more accessible, and dramatically more efficient.
Latest Videos FromTechRadar
But if our ability to generate code has increased so dramatically, why aren’t software organizations moving at the same pace?
General Manager of CX Engineering at NiCE.
AI-enabled development offers a more nuanced story. AI coding agents can handle work that previously required additional engineering capacity, from creating the foundations of an application to generating tests and accelerating implementation. A group of two or three people can try out an idea that might once have required a much larger team, more specialized skills, and considerably more time.
But speed is only part of the story. The more important change is that AI expands the ambition of each team.
A 2025 study by researchers at Tilburg University found that while AI coding tools increase output, they also increase the burden on senior developers. After analyzing thousands of open-source projects, the researchers found that experienced developers spent significantly more time reviewing AI-generated code, while their own coding output declined by nearly 20%. In other words, AI accelerates code creation, but it does not eliminate the need for human expertise to govern, validate, and maintain that code over time.
The implications extend well beyond senior developers. They force us to rethink what we actually measure.
More code doesn’t mean more progress
Excess code is only the symptom. The real challenge is keeping it under control. As codebases continue to grow, the question becomes even more pressing: if software can be built faster than ever before, why are organizations still struggling to move at the same pace? Because in enterprise software, writing code is only the beginning.
Every change must go through quality assurance, security reviews, regulatory compliance, and complex integration processes. Those steps cannot be accelerated at the same pace as code generation.
As AI speeds up development, what happens after the code is written becomes even more critical.
Organizations are measured not by how much code they produce, but by their ability to run stable, secure, and maintainable systems over time. That is why the defining challenge of the coming years will not be a shortage of developers, but managing complexity at scale.
AI changes the unit of innovation
AI coding agents are a major reason for this shift. They allow smaller groups to take on work that previously required a larger roster, without narrowing the ambition or depth of what they attempt.
That changes the unit of innovation. Progress is no longer determined only by how many people can be working on an idea.
The implications extend beyond individual developers. As AI changes the role of software engineers, team leaders need to consider not simply how quickly teams can produce code, but how effectively they can define problems, validate results and apply human judgement.
The real question is no longer how much software an organization can produce. It is how much of that software can be operated, maintained, and continuously evolved over time.
What happens after the code is written
The easier it becomes to generate code, the easier it becomes to generate complexity. Organizations can introduce new capabilities at unprecedented speed, but every additional component brings new requirements for testing, security, integration, monitoring, and long-term maintenance. AI shortens the path to writing software. It does not shorten the path to building reliable enterprise systems.
That is the new bottleneck in software development – not the ability to generate more code, but the ability to manage the complexity that code creates.
This is where governance, oversight and maintainability become increasingly important. A prototype is not a finished product, and it should not be treated as one. It still requires validation, product discipline, security, reliability, and a clear understanding of the customer or business need it serves.
The opportunity is to turn AI from a separate initiative into an everyday operating model for innovation, while maintaining the standards organizations expect around security, reliability, governance, and measurable business value.
This also means looking beyond traditional measures of development output. In a world where AI writes a growing share of the code, how many features were delivered, how many tasks were completed, and how quickly projects moved forward tell only part of the story. The ability to run stable, secure and maintainable systems over time matters just as much.
AI has solved one of software’s biggest challenges: generating more code in less time. Now the industry faces a harder question: how do we prevent speed from becoming technical debt?
We’ve featured the best AI tool.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit
