AI Is Already Part of the Dev Process

AI is no longer new in development, but a regular part of everyday work. In our process, it has proven effective to start with a solid analysis through OpenSpec, where we break down the requirements into smaller parts that can be verified, prioritized, and translated into concrete steps. Thanks to this, from the start we know what is critical, what is nice-to-have, and where it makes sense to apply automation. The analysis is therefore not a one-time document at the beginning of the project, but a living resource that the team continuously revisits.

In practice, we operate so that a human holds the direction, context, and final responsibility, while AI agents and subagents handle specialized tasks. One prepares implementation variants, another generates test scenarios, and another assists with documentation. This model significantly increases productivity because part of the work runs in parallel, yet quality remains under control. For project management, this is a big advantage: the plan is more predictable, the team continuously delivers outputs, and stakeholders can be shown concrete progress instead of general status updates.

Today, it is common for agents to generate proposals and code continuously 24/7. The team then selects the best variants, validates them, and moves them to release. The difference compared to the past is mainly in the pace and smoothness of iterations. At the same time, speed is not the only metric. We also monitor release stability, the number of rollbacks, documentation quality, and the impact of changes on operations. Without this discipline, a faster pace would not make long-term sense.

It works best where the roles of humans and agents are clearly divided. The human holds the direction and responsibility, the agents maintain performance and continuity of work. In such a setup, AI does not become a “flashy extra feature” but a natural part of the modern dev process.