In July, Evotix and the What Works Institute co-hosted a high-level roundtable, convening senior EHS leaders to discuss the evolving landscape of serious injury and fatality (SIF) prevention, artificial intelligence (AI) in safety, and the rise of human-centric risk strategies.

The What Works Institute is a new research-to-practice think tank and learning community for environmental, health, safety and sustainability.

Evotix tailors EHS and sustainability solutions with scalable software for managing incidents, audits, contractors, training, risks and hazards, regulatory compliance and other EHS&S applications. 

Theme 1: Rethinking SIFs: Definitions, Models, and Modularity

The ongoing lack of consensus around what constitutes a serious injury or fatality (SIF) is a significant challenge, participants said. Within their own organizations, definitions vary - ranging from high-impact physical incidents to long-latency exposures, and increasingly, to psychologically and emotionally disruptive events. Some organizations apply a strict physical severity lens, while others broaden the scope to include mental health consequences or life-altering diagnoses that may emerge over time.

Without a shared understanding of what qualifies as a SIF, efforts to align metrics, prioritize resources, or benchmark against peers is muddied. Several participants cited discrepancies between frontline language, corporate reporting standards, and industry definitions that create friction in incident classification and root cause learning.

The group acknowledged the value of emerging definitions offered by organizations like the Edison Electric Institute (EEI) and ASTM, which seek to standardize SIF definitions. But a modular, customizable approach - one that allows organizations to align with broader standards while tailoring risk criteria to their operational realities - is seen as essential.

Maturity is a critical framing device. Early-stage organizations tend to focus on physical prevention — “arms and legs” safety — while those at more advanced stages are beginning to incorporate contributing factors like time pressure, decision complexity, or emotional strain.

Leaders expressed a need to both elevate the clarity of serious risk definitions and retain the agility to evolve those definitions as science, data, and frontline insights continue to emerge.

Theme 2: Human-Centric EHS in Practice

Many participants described active efforts to move beyond compliance-driven programs toward more integrated human-centric approaches. The interlocking concepts of psychological safety, psychosocial risk, and cognitive inclusion are all seen as increasingly essential to understanding and mitigating serious harm.

A human-centric model accepts that perception, behavior, and response are shaped by stress, fatigue, neurodiversity, and emotional dynamics — factors that, though less visible than obvious physical hazards, are no less consequential. The group discussed examples of integrating these factors into tools, language, and leadership behaviors, such as revising investigation protocols to assess emotional and cognitive load as contributors to incidents, not just physical conditions or procedural breakdowns. Incorporating psychological safety indicators into safety climate assessments and designing signage and procedures to reflect cognitive diversity on the shop floor are other examples.

The discussion underscored that human-centricity isn’t a separate initiative - it is an organizing principle for how modern safety systems must evolve.

Theme 3: Cultural Conditions

Leaders repeatedly emphasized that no framework or tool can deliver sustained impact without an environment in which workers feel safe to speak up and leadership is visibly aligned. Systems that tie executive bonuses or ESG performance to metrics like TRIR or SIF count can unintentionally suppress reporting, distort focus, or drive performance theater. The group discussed the importance of designing incentives that reward learning, curiosity, and meaningful prevention activity, rather than simply the absence of recorded harm.

Participants noted that some of their most accurate and actionable SIF risk indicators come not from dashboards, but from direct questions posed to workers: “Where do you think the next serious incident will happen?” or “What’s most likely to hurt someone here?” These candid, often informal exchanges surface patterns and blind spots that formal systems often miss.

Many organizations have layered programs involving safety culture, wellbeing, SIF, behavioral observation that lack clear integration. Several attendees advocated for a return to fundamentals: fewer, clearer systems with well-communicated purpose and built in partnership with those who use them.

Theme 4: AI in Safety

Several organizations reported early-stage experimentation with AI-generated toolbox talks, report summarization, or analytics dashboards that surface leading indicators. Participants cautioned that “garbage in, garbage out” remains a pressing concern — highlighting the need for clean, context-rich, and representative data.

Most participants placed their organizations at an early stage along the AI maturity spectrum, between ad hoc use and strategic exploration. A small number described more integrated models, where AI tools are tied to enterprise safety systems, though these remained the exception.

Participants agreed that AI tools must amplify human judgment, not bypass it — and any system introduced must reinforce, rather than erode, the trust that underpins effective safety culture.

Theme 5: Metrics

Participants voiced a clear and shared frustration: the Total Recordable Incident Rate (TRIR) continues to dominate dashboards, often serving as a proxy for safety performance, even as its limitations in preventing serious harm are widely acknowledged. Several attendees noted that their organizations still rely heavily on lagging indicators, with little integration of metrics tied to exposure, psychosocial risk, or system health. This is a critical barrier to shifting from reactive compliance to proactive prevention.

Shifting mindsets around measurement, especially among executives and auditors, remains a slow and political process, participants acknowledged. Ideas for better metrics are emerging but implementation is lagging. The group underscored the need to redesign what safety success looks like, and to build measurement systems matching the complexity of risk environments.

The roundtable also included suggested topics to guide refinement of the What Works Institute/Evotix 2025 EHS Strategy & Innovation Survey instrument. There is interest in using the survey as a catalyst for peer conversation, reflection, and future collaboration.

The insights captured in this roundtable will directly inform the next iteration of the What Works Institute Evotix Insights Report and the accompanying 2025 survey. The group affirmed that the momentum for change in EHS&S is real, demanding collaboration, experimentation and honest reflection, and the field is ready for the work ahead.