AI adoption succeeds in EHS when priorities come before the technology

Most failed AI pilots in EHS don’t fail because the technology doesn’t work. They fail because the technology was chosen first, and the business case was built afterward. A tool gets selected because it’s interesting or because a vendor pitched it well, and then the team discovers there’s no urgent internal problem it’s solving — which means there’s no budget owner, no executive sponsor, and no reason for adoption to survive past the pilot stage.

Porter’s alternative is straightforward but easy to skip under pressure to “do something with AI”: identify the two or three problems already costing the EHS program the most time or risk — inconsistent incident data, slow investigations, sustainability reporting overhead — and evaluate AI against those specific problems. That ordering matters more than it sounds. A tool tied to a named priority has a sponsor built in. A tool chosen for its novelty has to go find one.

The same discipline applies to trust, which Porter frames as something built in layers rather than granted upfront. Education comes first — understanding what a tool actually does before deciding whether to rely on it. Evidence comes next, in the form of demonstrations or documented results. But the layer that actually shifts behavior is the third one: a peer describing a concrete outcome in their own words.