Key Highlights
- Traditional safety metrics often miss the human factors contributing to serious harm, emphasizing the need for human-centric approaches that address stress, fatigue, and psychological risks.
- Integrating psychosocial risk assessments and redesigning investigations can help organizations identify and mitigate root causes of serious injuries.
- AI has the potential to revolutionize safety management, but requires robust governance, human oversight, and cross-disciplinary collaboration to prevent risks and biases.
- A systemic rebuild involving clear definitions, integrated metrics, and responsible AI adoption is essential to change the trajectory of serious injuries and fatalities by 2026.
Despite decades of effort, serious injury & fatality (SIF) rates remain alarmingly steady. This isn’t from a lack of commitment but a misalignment in how risk is defined, measured and managed. A unique opportunity to radically reduce SIFs is emerging, but it will require a system rebuild, not more add-ons.
When We Disagree What Serious Means
While basic compliance and incident-prevention programs have driven down minor-incident rates in past decades, traditional safety efforts have yielded diminishing returns on serious work injuries. The good news is that organizations are now pushing beyond compliance checklists to focus on life-altering harm. According to the Risk Recalibrated: the 2026 Executive Leadership Report, 80% of responding organizations have SIF prevention programs in place. But the lack of universal acceptance of what constitutes a SIF is a problem.
SIF definitions vary not only across companies but sometimes within them. Some might restrict SIFs to immediate, high-severity physical trauma, while others include long illnesses or psychological trauma—to cite two common examples. These widely varying definitions lead to inconsistent classifications, uneven data, complications in operational benchmarking and therefore, confused priorities. Nearly 1 in 5 EHS leaders say traditional safety metrics have no relation to real risk, and more than half say they only partially reflect SIF drivers.
This misalignment results in critical risk exposure and a disconnect between executives, EHS teams and frontline workers, which only perpetuates the problem.
To make meaningful progress in reducing serious harm, alignment may not require a universal definition, but it does demand internal clarity. The first step in lowering SIF rates is for organizations to adopt their own definitions. You can start this process by convening key, cross-functional stakeholders, including safety teams, operations leaders and HR, to define SIF clearly across the organization.
Serious Harm Rarely Starts with a Single Incident
Even with aligned definitions, organizations will still struggle to reduce serious harm if what they measure is disconnected from how harm forms.
Traditional safety strategies focus heavily on achieving compliance and tracking errors. This approach fails to deliver real improvement in SIF rates because serious harm rarely starts with a single unsafe act. More often, it starts with someone working under pressure. All people face real-world pressure, like stress, fatigue, burnout, grief and literacy/language barriers, and coping mechanisms vary. Conventional indicators miss this reality entirely.
Human-centric workplaces rely on strategies that prioritize employee well-being and flexibility and encourage design processes and workflows that account for everyday human pressures. Blending these strategies with EHS will provide a greater understanding of potential serious injury or fatality (PSIF), and the data show that organizations are increasingly acknowledging this, calling out control of work conditions (71%), mental health strain (66%), and fatigue (60%) as indicators.
But recognition is outpacing integration. Most (89%) say the human-factor contributors are not yet integrated into safety or SIF strategies.
Prevention-focused Metrics
Recognition alone won’t reduce serious harm. Closing the knowing-doing gap requires integration and simplification, not more parallel initiatives. This includes drilling down into key metrics and redefining events—such as revamping investigations to account for individuals’ emotional and cognitive states, and adapting training and signage for different literacy styles.
Other examples might include tracking high-potential near-misses or the percentage of inspections that include a psychosocial risk observation, breaking the silos among EHS, HR/occupational health, operations and IT to co-sponsor efforts that cover topics such as fatigue management, workload redesign and supervisory skills in empathetic leadership.
A practical, shared maturity model or tool for psychosocial-risk integration can map a path from reactive, post-incident support to embedded, human considerations across design reviews, pre-job planning, investigations and KPIs. This is how organizations prioritize high-leverage changes, pilot, measure and scale what works.
The key to SIF reduction is measuring what prevents fatalities and life-altering harm, not counting after-the-fact incidents.
AI Governance is Make-or-Break
As organizations work to integrate and scale prevention metrics and tactics, many are turning to AI—42% are now piloting AI, and another 32% are exploring use cases. And for good reason.
AI offers new ways to predict, prevent and respond to risk in ways humans struggle to scale manually. Emerging solutions have the potential to detect potentially unsafe conditions, capture and classify incidents, support audits and inspections by consolidating and summarizing findings, and connect information across incidents, inspections, audits and near misses that may point to emerging risks.
The potential for AI to measurably reduce safety risk is undoubtedly high, but so are the stakes of getting it wrong. This is where strong governance is emerging as make-or-break. EHS teams have warranted concerns over data quality and bias (58%), privacy (39%) and a lack of explainability (36%).
Clear guardrails are needed to prevent missing key controls, for example, or recommending an unsafe process. The single most important guardrail is human input and approval. This includes cross-functional partners, including IT, legal, operations and worker representatives who engage early to set guardrails on data permissions and anonymization, model validation against known cases, monitoring for drift and bias, and establishing clear boundaries on use.
Changing the SIF Trajectory
Meaningful SIF reduction is possible in 2026, but small steps won’t get us there. It’s time for a complete system rebuild. This means aligning definitions, understanding, integrating and measuring root risk cause and responsibly adopting AI for scale. Organizations that take this approach will see safer workplaces and play an important role in changing the trajectory of serious injuries and fatalities.
About the Author

Jonathan English
CEO
Jonathan English is CEO of Evotix, a global provider of environmental, health, safety and sustainability (EHS&S) solutions.
