A worker can be exposed to a chemical today and develop a serious illness years later. Many SIF frameworks are better at recognizing an immediate injury than that kind of occupational health risk. When the same exposure can count as a potential SIF under one model and not another, we need to ask whether our definitions are helping us prevent serious harm or leaving some of it out of view

It’s easy to measure whether something was completed. The harder and more useful question is whether it actually made a difference. An inspection, audit or action being closed doesn’t necessarily tell you whether the risk has changed. We need measures that help us understand whether the controls we’ve put in place are actually working.

Agentic AI is coming to EHS, and it will be transformative. But trust has to be earned, and in safety the cost of getting it wrong is measured in people. For now, the human in the loop is essential, and that person needs the expertise, context and authority to challenge or override what the AI recommends. Oversight only adds value when someone is in a position to exercise real judgment, not just confirm what the technology produced.

As EHS leaders, we’re always looking for ways to improve, but improvement doesn’t have to mean adding another program, process or tool. Sometimes the better question is what we can simplify. The strongest systems give people clear expectations, useful information and processes they can actually follow while still evolving as the business changes.

We can put sophisticated tools in front of a new EHS professional, but that doesn’t give them the experience to know what they’re looking at. Some arrive with little exposure to SIF, human performance or psychosocial risk, then step into complex operations and mature systems. We need to think much more deliberately about what they learn in their first few years, and who is there to teach them.