The next revolution in software development: AI reviewing its own code

AI has not only made coding faster, but it has also significantly increased the amount of code. Before the AI era, a single developer would write code in their own branch, after which a colleague would conduct a review and check whether everything had been done according to the goals. 

Because AI has made code generation so fast and cheap, many are facing a new situation: there is more code than any organisation, team, or specialist could ever check manually. For this reason, an inevitable evolutionary step is that AI is also needed in this phase of work. 

The next essential developmental leap in AI-driven software development is, therefore, to largely automate software quality assurance completely. Otherwise, we face an era of sub-standard software, and at the same time, the benefits offered by AI will be cut short.

This is a factor that will largely determine which companies succeed and which do not. Few of us would think to demand at a car dealership that a human has checked every single bolt on the vehicle with their own eyes. We trust automated testing, and exactly the same will happen to code. In other words, the same technology that created the original problem—the explosive growth of code – will also solve it. What is important in this AI-driven developmental leap as well is that we build working processes and ensure trust in the system.

In this transition, humans will become a kind of "supervisor machine operator". While the machine handles verification – checking, for example, whether the code was written according to the given specifications and guidelines – top experts are needed more than ever to validate the quality of the work. For instance, specialists can check if the end result and the user interface are actually easy to use, fit for purpose, and meet user needs. Understanding the big picture will therefore remain firmly within the human domain, at least for now.

At the same time, the professional focus of the specialist will shift even more from technical execution towards client consulting and direct client interaction. One must understand business in order to explain to the AI what needs to be solved and within what constraints.

When basic applications can be mass-produced with sufficient quality, the true value will increasingly arise from human innovation and problem-solving. AI does not eliminate the challenges of IT and software development; it merely changes their nature. The bottleneck in the industry will no longer be the actual implementation, but the ability to conceptualise and define precisely what we want the machine to do. Those who cling solely to "artisan coding" and manual code reviews will inevitably be left behind in development.

Juha Huttunen, CEO, Rakettitiede

This article was published in Tivi in July 2026.