The rapid development of artificial intelligence technologies is dictating new rules for working with information. While just six months ago, the limit of automation in Health, Safety, and Environment (HSE) was considered to be writing prompts for language models to generate text, today the focus has shifted toward autonomous systems. AI agents and the concept of "vibecoding" are opening up opportunities where a specialist can not only delegate routine tasks to a neural network but also independently create corporate software without programming skills.
Large businesses are already actively integrating artificial intelligence into their processes. Companies in the metallurgical and mining sectors are creating their own industry-specific language models and implementing RAG (Retrieval-Augmented Generation) systems to work with vast arrays of regulatory documentation. However, the main technological shift of recent months has been the transition from simple assistants to AI agents.
Unlike basic neural networks, an agent is capable of executing complex, multi-step chains of tasks autonomously. During his presentation, Rinat Fatkhutdinov demonstrates the operation of a local agent that takes over the full cycle of risk assessment. The process looks like this: the system receives a photograph of a workplace, visually identifies hazards in accordance with Order No. 776n, automatically opens Excel on the user's computer, fills out the risk register, calculates the necessary budget for corrective actions, and generates a final presentation for management. Human involvement is reduced solely to approving key steps.
Another significant trend is vibecoding. This is an approach where the user describes the program's logic in natural language, and artificial intelligence writes and assembles the working code. For HSE specialists, this means independence from long IT development cycles.
The speaker analyzes this approach using his own experience in creating a comprehensive management system for eight types of safety (including fire, environmental, and industrial). With the help of AI, an MVP (Minimum Viable Product) was developed that automatically builds the organizational structure of the enterprise, assigns roles, and generates local regulatory acts. For example, the system is capable of independently drafting fire safety instructions, taking into account the specifics of a particular building, the location of smoking areas, and designated responsible persons. Developing such a tool using traditional methods would have required months of work by an entire team of programmers, whereas vibecoding allowed the project to be implemented by a single subject matter expert in just a few weeks.
The implementation of autonomous systems changes not only processes but also competency requirements. There is a growing need for "knowledge brokers"—specialists who deeply understand industrial safety methodology while simultaneously mastering AI tools. They become the vital link between the real needs of production and IT departments.
Artificial intelligence does not replace the expert, but it radically changes the value of their work. Being freed from routine paperwork allows them to focus on proactive actions, the development of a safety culture, and strategic risk management.