In recent months, concerns about the continuously increasing volume of code reviews have become more frequent among engineering leaders. This trend has emerged as a key topic because, with the growing use of advanced artificial intelligence code generation tools, the existing bottleneck in the software development cycle has shifted from the code writing phase to the code review phase.
In response to this situation, a number of AI-powered tools are emerging and developing in the market, aiming to help manage the growing burden of reviews. These tools are being intensively tested and adopted to automate and streamline the code inspection process.
Planning a new website or app?
Tell me what you have in mind — I will take a look and advise on the best way forward.
In parallel, many larger companies are investing in developing their own internal solutions to improve the code review experience. Typical features of these custom systems include intelligent reviewer assignment, which optimizes task distribution, and change risk profiling, which estimates the potential impact of modifications and alerts developers to critical points requiring increased attention. It has been observed that for these companies, internal implementations often prove more effective than integrating external vendors.
Another line of thought focuses on code verification instead of mere review. Although seemingly easier said than done, theoretically, thorough testing should ensure that the code functions as expected. However, the question remains what exactly constitutes "through" testing – what types of tests (unit, integration, end-to-end, fuzz testing, formal methods) are needed, and how should they be connected with the overall system observability? And how can one ensure that newly added tests genuinely verify the desired functionality?
The excessive burden of code reviews leads to engineer burnout, often resulting in superficial or low-quality reviews. It is frequently heard that developers approve AI-generated code simply because automated tools found no obvious issues, without delving deeper into the essence of the changes. On the other hand, those who dedicate the same effort and energy to reviews as before feel overwhelmed by the increasing number of merge requests, many of which are a result of automatic generation. The problems are clear, but current solutions appear to be more experimental at this stage.