AI for Health, Safety & Environment at the workplace

Image

Reducing our lost time injury frequency rate 
 

Since 2016, Boots on Ground (BoG) has helped reduce our Lost Time Injury Frequency Rate by 63%, reflecting our conscious shift to instill a culture of genuine commitment and proactive ownership – engaging 65 000 employees and contractors worldwide.

As the platform has grown to 16 500 active users and 450 000 coaching sessions, the richness of the data it generates has improved. The question we faced was how to turn this high volume of data into meaningful, timely decisions at multiple sites, simultaneously? The answer was AI.

Image

Using AI to predict serious incidents
 

Every month, across over 3 000 sites in 45 countries, Predict Au tool uses an AI model to analyze over 1 500 data points – from leadership presence in the field to the speed at which hazards are closed – flagging the sites most likely to experience incidents before they happen.

Just as an email spam filter learns the patterns of unwanted messages and flags them before they reach your inbox, Holcim’s Predict Au learns the patterns that can precede serious incidents onsite, and raises the alarm early enough for us to act.

The model evaluates every Holcim site, drawing on data from BoG and iCare – Holcim’s global integrated digital platform to report and manage HSE data – to analyze over 10 million data patterns. It tracks time spent in the field, past incident history, coaching frequency, hazard trends, action closure rates and critical risk recurrence.

Strong third-party validation 
 

Image

It scores sites, not people, so no personal data is used. And it does not make decisions. The model shows where to look and then leaders and teams decide what to do. A flagged site receives immediate, structured support from HSE professionals and site leadership builds an action plan. The site is then progressively de-escalated as its risk indicators improve until it returns to normal status.

Predict Au has been independently validated by third-party AI and machine learning experts, confirming its correlations between leading indicators and actual safety outcomes. In practice it has flagged sites five to ten months ahead of serious incidents, and where teams acted on the alert by strengthening field presence and critical controls, those sites returned to normal with zero critical incidents.

When a site is flagged, there is always a human in the loop: Holcim's HSE professionals and leaders lead on the ground, deploying their expertise where the data shows it is needed most, where AI helps us to cut through the noise and focus on actions that ensure safety.

A “Fitness Score” for Holcim sites
 

Alongside Predict Au, Holcim has introduced a “Fitness Score” for each of our sites, a single KPI that rates the site's HSE program in one clear, comparable metric. It integrates traditional leading and lagging indicators with the real-time risk signals detected by Predict Au, giving leaders an instant read on where a site stands and, crucially, where it needs to go. This Fitness Score ensures that no strong country average can mask an underperforming site: elevated risk at a single site pulls the entire country's score down, ensuring risk is always visible.

Both Predict Au and the Fitness Score represent an industry-first approach to predictive HSE management at scale: real-time AI-driven risk assessment deployed simultaneously across thousands of sites, built on a strategy that puts field intelligence in the hands of leadership at every level - from plant managers to our Executive Committee.

Building on this foundation, Predict 2.0 is already in development, designed to go further, reach deeper, and extend the full power of AI machine learning to every Holcim site.

At Holcim, AI is not replacing the human side of HSE. It is giving it back: fewer hours reading data, more hours in the field with the people we are there to protect. Technology does the reading. Our people do the preventing.