AI-Assisted Software Development in Practice at KINALI

Why this matters now

Using AI just for the sake of it does not mean it automatically brings value. What matters more is where it actually saves time, improves code quality, and helps prevent problems before they move further into development.

That is the perspective we care about at KINALI. Not AI as a trend, but AI as a practical tool that assists with software development as well as other areas of our business.

How KINALI uses AI today

We have been using ChatGPT Business since around mid-2024. We have used Codex almost since its release in May 2025, and Copilot slightly longer than that.

Since December 2025, we have also had regular meetings to share experiences with AI use. The goal is simple: compare what works, where it helps, and how to apply it in a way that supports and enhances our work.

The clearest measured example so far

The strongest measured example at the moment comes from the Mycronic Line Programming team.

Based on the team’s current experience, AI-supported development is helping deliver roughly 35% higher team velocity with the same number of people.

A large part of that comes from AI-supported code review. According to Honza Uhlíř, the effect is not only higher productivity. In many cases, a major benefit is better code quality. AI helps catch bugs, performance issues, and code smells earlier, which saves time in review and reduces the chance that these issues turn into larger problems later.

In more than 50% of reviewed cases, AI feedback led to a correction.

How the review workflow works

In Honza Uhlíř’s workflow, the AI review tool opens threads automatically when it finds a bug, performance issue, code smell, or similar technical problem. It also proposes a fix.

Then it returns to the threads it already opened and checks whether the code has changed or whether the developer has responded. If the issue has been resolved or the developer’s justification is valid, the tool closes the thread automatically.

So far, the tool has opened 242 AI review threads across 71 merge requests. Out of those, 209 were resolved and 33 were still open at the time of reporting.

In 56% of cases, the developer fixed the code and the bot verified the fix. In another 10%, the developer justified the solution and the bot independently verified and accepted it. Overall, 66% of resolved threads were closed autonomously by the AI agent.

This reduces manual review overhead, saves time for developers, and helps keep review attention focused on broader feature and architecture questions.

Beyond one team

The Mycronic Line Programming team is the clearest measured case, but it is not the only place where AI is helping at KINALI.

Across other teams at KINALI, we also use AI for repetitive tasks, bug detection, review support, merge request review, and smaller programming tasks. Jiří Zajíc, for example, uses AI and Codex in merge request review and in some cases in programming itself.

This saves time on simpler work and helps surface technical issues earlier. That is why, for us, AI is not only a productivity topic. It is also a software quality topic.

AI beyond customer development

We also use AI outside direct customer software development.

This includes internal tools for candidate logging and processing, invoice management, and their automatic conversion from PDF documents into the correct format for accounting workflows, project planning and budgeting, and attendance and presence tracking. Almost 100% of the code in these tools was written using AI.

These cases show that the value is not limited to one team or one workflow. AI can also help speed up internal application development and reduce manual operational work.

Conclusion

Using AI is important to us at KINALI. Not because it sounds progressive, but because it helps us deliver software faster and with even higher quality, and also helps us increase efficiency in other areas like project planning, HR, and other key business processes.