The study showed the share of AI agents in creating models

A research group analyzed 769 task logs when creating their own AI model to assess the actual contribution of autonomous systems to the development process.

According to the study results from research, AI agents generate up to 55% of methodology proposals, but humans still control the process and make over 85% of the final decisions.

The analysis showed that about a third of all tasks performed, engineers would not even have attempted to solve without the help of specialized algorithms, the authors of the study emphasize that the growth of agent activity in projects does not mean the automation of key decision-making.

The data obtained demonstrate the current boundaries of autonomy in machine learning engineering, where AI takes on routine and hypothesis generation, and final responsibility remains with humans.