Why clean data and human judgment matter more than the algorithm
Many years ago, a number of organizations managed incident prevention in the rear-view mirror. Something happened, it was documented, corrective actions were completed, and the data went into a spreadsheet or a paper file that nobody revisited. According to BIS Safety Software, the Sherwood Park, Alberta company that builds environment, health, and safety software for many organizations across the country, that is no longer good enough for the safety leaders it works with.
"What we have seen more recently is a genuine appetite to get ahead of incidents rather than just respond to them," says Dan MacDonald, BIS Safety Software's founder and CEO. "Organizations are asking better questions of their data: not just how many incidents happened, but where are they occurring, what hazards or high-risk activities are driving them, are the same trends appearing across sites, are training gaps showing up in the same locations as incidents?"
MacDonald says Canadian organizations, particularly in energy and construction, have been at the forefront of that evolution "because the regulatory environment and the cost of incidents in those sectors makes the business case very clear."
The misconception that trips organizations up
The single most common misunderstanding, in MacDonald's experience, is that AI will replace the people who run a safety program.
"Organizations sometimes approach these tools expecting a kind of oracle that tells them what to do, and when the reality turns out to be a system that surfaces patterns and automates repetitive tasks rather than making decisions for people, there is occasionally a gap between expectation and experience," MacDonald says. "Our job is to reframe AI as acceleration, not replacement. The platform removes the manual work so that the safety manager can focus their expertise on interpretation and action rather than data entry and report generation."
A second misconception, MacDonald says, is that AI tools need perfect data to be useful. In practice, "organizations can start getting value from analytics even when their historical data is incomplete, because the system begins building a cleaner picture from the moment they start using it properly."
That plays out in real cases. MacDonald points to clients in the energy sector where certain certifications are mandatory before a worker can access a site. "Our system automatically flags workers whose certifications are approaching expiry and can block or restrict site access for individuals who are no longer compliant," he says. "Without that automation, the gap between expiry and renewal often goes unnoticed until an audit or, worse, until something goes wrong."
Benefits beyond the incident log
The gains organizations report are as much cultural as operational. Centralizing safety data reduces the administrative time safety teams spend on compliance, sometimes dramatically, and changes how people relate to safety when records are visible rather than buried in binders.
"Frontline workers take more ownership when they can see their own compliance status," MacDonald says. "Supervisors make better decisions when they have real-time visibility in their team's readiness. Leadership teams develop more confidence in the program when they can pull accurate, auditable data on demand." Collectively, he says, those changes move organizations "from treating safety as a cost center to treating it as an operational advantage."
On the question of responsible AI, MacDonald draws a firm line between support and surveillance. "A system that flags a training gap is doing its job well," he says. "A system that uses that same data to score or rank individual workers in ways they have not consented to would cross a line." MacDonald says the company maintains permission and access levels so sensitive data is visible only to those who need it and designs its tools "to support accountability rather than surveillance."
Looking ahead, MacDonald sees the field moving "from descriptive analytics, which tells you what occurred, through diagnostic analytics, which tells you why, toward predictive and prescriptive tools that tell you what is likely to happen and what to do about it." His advice to Canadian safety leaders preparing for that shift is unglamorous but pointed: get your data in order now and start building AI literacy across the team.
"Organizations that have invested in digital forms, consistent incident reporting, and structured training records will be positioned to take full advantage of predictive tools as they mature," MacDonald says. "Those who are still managing compliance on spreadsheets or paper will find the gap increasingly difficult to close."
