"Safety Pill" — Engaging Workers in Risk Analysis Through AI Workplace Visualizations

From Risk "Blindness" to Conscious Safety

In modern HSE, one of the main problems remains "blurring" of vision: workers stop noticing everyday hazards in their familiar workplaces. Traditional briefings and showing real photos often fail to elicit the proper response, as the brain automatically ignores a familiar picture. During the webinar, Svetlana Kautova analyzes an unconventional approach to solving this problem — using artificial intelligence to create personalized workplace visualizations with intentionally embedded risks.

Mechanics of AI Visualization: How the Method Works

The speaker demonstrates, using neural networks (specifically, the Leonardo platform), how realistic images of workshops, warehouses, and specific machines are generated. Typical violations are purposefully added to these images: lack of PPE, spilled liquids, unsecured loads, or poor ergonomics.

The essence of the approach lies in holding short 15-minute meetings directly in the departments. Workers are asked to analyze printed images and find the embedded hazards. This shifts the employee from being a passive subject of a briefing to the role of an active analyst, stimulating critical thinking and forming a habit of noticing deviations.

Why Generated Images are More Effective Than Real Photos

The presentation details the psychological aspect of perceiving visual information. Using real photos of workplaces often triggers a defensive reaction, is perceived as a witch hunt, or violates employees' personal boundaries. In addition, a completely familiar environment reduces concentration.

Generated images solve these problems:

• Cognitive dissonance: the worker sees a similar but slightly altered workplace, which forces the brain to engage and start looking for inconsistencies.
• Psychological safety: the absence of real people in the photo eliminates the fear of punishment and legal risks, turning the process into an educational game format.
• Modeling flexibility: AI allows adding any combination of risks, changing lighting and conditions, training personnel to anticipate hidden threats.

Implementation Results and Feedback Loop

An important element of the system is gamification and continuous dialogue. Workers not only find the embedded risks but also propose their own hazard scenarios via QR forms or questionnaires. These suggestions are then integrated into new AI generations. Over 5 months of the pilot project, employees identified more than 200 unique risks, most of which related specifically to behavioral habits.

This approach has not only increased personnel engagement but also promptly eliminated a number of real threats in production — from installing cable channels to purchasing tool stands, based solely on grassroots initiative.

What You Will Learn From This Webinar:

• How to properly write prompts for a neural network to get technically accurate images of production areas?
• Why is using real photos of violations less effective for engaging personnel than AI generations?
• How to organize feedback collection from workers and integrate their suggestions into new visual materials?
• What typical errors occur when generating workplace images and how to quickly correct them?
• How to introduce gamification elements into the occupational risk analysis process without using complex IT solutions?