How Cam Stevens Thinks About Technology as a Safety Signal

Episode 15 of Headline Podcast, published on March 18, 2026, brought Cam Stevens, CEO of PKG, into conversation with Andreza Araujo and Dr. Megan Tranter about safety technology, worker voice and the changing psychosocial risk profile of work. Stevens defended a clear thesis on the show: technology should answer a defined work problem, because a catalog of tools can be dangerous when leaders have not decided what needs to change.

That argument is more demanding than a technology review. It asks the EHS leader to examine how a tool changes attention, autonomy, reporting, supervision and the pace of decisions, which is why this episode belongs in the psychosocial-risks category as much as in safety technology.

Key Takeaways

Cam Stevens's central message is that technology selection begins with diagnosis. The organization should name the exposure, the decision that is failing and the evidence that would show improvement before it compares vendors or features. This matters because digital tools can reduce physical exposure while creating new pressure through monitoring, alerts, work intensification or uncertain data use. A useful technology decision therefore has two tests. It must improve the control in question, and it must leave people more able to see, question and manage risk.

• Define the work problem before reviewing the technology catalog.
• Separate useful signal from prediction hype and dashboard volume.
• Assess how the tool changes autonomy, monitoring and work pace.
• Protect worker voice when collecting operational or behavioral data.
• Review results after 30 days against the original risk question.

Why a technology decision is also a psychosocial-risk decision

Psychosocial risk is shaped by how work is designed, managed and experienced. A safety tool can therefore become part of the hazard picture when it changes how much control workers have, how quickly they must respond, how closely they feel watched or how safe it feels to challenge an automated recommendation. The WHO describes mental-health risks at work as connected to job content, work schedules, workplace characteristics and opportunities for development. Technology can alter each of those conditions.

The ILO and WHO policy brief on mental health at work in the era of digitalization identifies the pace of digital change as a factor that requires attention from leaders. That does not mean technology is inherently harmful. It means that a new platform, wearable, camera, algorithm or voice channel should enter the risk assessment rather than bypass it.

ISO 45003:2021 gives organizations guidance for managing psychosocial risk inside an occupational health and safety management system based on ISO 45001. Its relevance to this episode is practical. The tool is not the control by itself. The control is the work arrangement, decision path and leadership response that the tool helps people execute.

For a plant manager, that changes the approval question from “Does this system have enough features?” to “Which exposure will be lower, for whom, and what new pressure could the system introduce?”

Cam Stevens's first argument is problem-first adoption

Stevens said, “If we're very clear on the problem to solve, then a technology catalog is excellent. If we're not clear on the problem to solve, a technology catalog is very dangerous.” The quote is a direct challenge to feature-led procurement. A tool should be selected only after the operation can describe the current failure, the user who must act and the decision that should become easier or more reliable.

A practical problem statement has at least four parts. It identifies the exposure and the work situation in which it appears. It names the weak barrier and the response that must happen next. “We need better visibility” is not enough. “Maintenance planners cannot see that temporary changes have removed a critical control before the job starts” gives leaders a decision to test.

That distinction protects the organization from a familiar trap. When a system produces thousands of observations, alerts or voice fragments, the team can confuse activity with control. Data may be abundant while ownership remains absent.

Andreza Araujo's work across more than 250 cultural transformation projects reinforces this point from the culture side. The field does not become safer because a platform exists. It becomes safer when leaders make the expected action clear, support the person who raises the concern and close the loop visibly.

Prediction is not the same as a useful signal

The episode distinguishes between technology that helps people notice a condition and technology that claims to predict an incident. Those are different promises. A signal can support attention and a decision. A prediction claim can create false confidence, especially when leaders do not understand the model, the data boundary or the conditions under which it fails. The safer question is whether the tool improves a defined decision at a defined point in the work.

Suppose a system flags fatigue, congestion or repeated deviations. The leader should ask what happens within 24 hours of the flag. Who reviews it? What information is added by the worker? Which control can change? What evidence shows that the response reached the field? If those questions have no owner, the alert is only a new queue.

James Reason's distinction between active failures and latent conditions remains useful here. A late response or missed warning may be visible at the sharp end, but the organization may have created the conditions through poor interface design, ambiguous thresholds, workload, weak supervision or unclear escalation.

Technology should make those conditions easier to discuss, not turn them into a score that replaces judgment. A predictive label that cannot be challenged may suppress the very information the organization needs.

Voice technology can widen access to information

Stevens described voice technology as a possible source of operational information, while warning that technology should not be used simply because it exists. Voice can lower the effort required to report a weak signal, especially when typing is impractical or when the worker has a richer account than a checkbox allows. It can also feel invasive if people do not know whether the system is listening, how data is stored or whether a concern will be used against them.

OSHA's Safe and Sound guidance places management leadership, worker participation and a systematic approach to finding and fixing hazards at the center of a safety program. A voice channel can support worker participation when it makes it easier to share experience and receive a response. It cannot create participation if the underlying response is dismissive.

The ILO working paper on AI systems at work identifies surveillance, changes in job autonomy and data-driven management as emerging dimensions of the psychosocial work environment. Those risks should be tested before launch, not discovered after trust has already declined.

Leaders should state four boundaries in plain language. They should explain what is collected, what is not collected, who can access it and what action follows a report. If the answer changes by team or country, that difference belongs in the rollout plan.

Technology can elevate the human experience, but not everywhere

Stevens said, “We can elevate the human experience with technology, but there are certainly times when technology should absolutely not be used.” That boundary is the episode's most important protection against enthusiasm. A tool is unsuitable when it replaces a conversation that requires context, increases fear without improving control or shifts responsibility to a worker who cannot change the conditions around the task.

Use technology when it improves access to critical information, removes avoidable physical exposure, supports a timely decision or helps workers communicate a condition that leaders cannot otherwise see. Pause when it adds observation without response, measures behavior without context or makes a person responsible for a risk created by staffing, design or scheduling.

The decision also depends on proportionality. A low-consequence administrative delay should not receive the same surveillance architecture as a critical control. A system that treats every deviation as equal can make the important signal harder to recognize.

This is where *Antifragile Leadership* offers a useful brand-adjacent lens. Leaders should design systems that become more informative after friction, rather than systems that punish the person who reveals friction. The test is whether a concern produces better conditions or merely a better record.

Before and after a problem-first technology review

The difference between feature-led adoption and problem-first adoption is visible in the questions leaders ask, the evidence they collect and the response they expect. The comparison below translates the episode into a decision tool for an EHS manager who has 30 days to assess a proposed platform or digital workflow.

The second column is not anti-technology. It is anti-ambiguity. When the original problem is explicit, the organization can decide whether a tool is the right answer, whether a work redesign is better or whether a conversation is the only proportionate intervention.

Recommendation

Use Cam Stevens's problem-first test before approving any safety technology. Write one sentence that names the exposure, the work context, the weak barrier and the decision that should improve. Then run a 30-day review that includes workers from the affected team, the control owner, the supervisor and the person responsible for psychosocial-risk assessment.

During the first 7 days, document the baseline and explain the data boundaries. During the next 14 days, review actual use, worker concerns, response time and any new pressure created by the tool. In the final 9 days, compare the evidence with the original problem statement and decide whether to continue, redesign or stop.

The review should answer four questions. Did the tool make the critical decision clearer? Did a worker gain a safer route to raise information? Did leaders respond within the promised 24 hours? Did the technology reduce exposure without transferring hidden pressure to the person closest to the task?

For a broader psychosocial-risk workflow, connect this review to the industrial-plant escalation workflow, the psychosocial risk appetite guide and the ethics and trust decision model. Those internal references help place technology inside governance rather than treating it as a standalone purchase.

FAQ

The episode does not offer a universal technology list. It offers a way to think. The most reliable choice depends on the exposure, the people who must act, the information they can access and the response leaders are prepared to make visible.

What should leaders do when a vendor promises prediction? Ask which decision the prediction improves, what data the model needs, where it can fail and how a worker can challenge it. A prediction that cannot change a control is not a safety intervention.

Is voice technology suitable for every operation? No. It may help where workers need a low-friction reporting route, but it should not be used where privacy, consent, context or response ownership cannot be protected.

How does ISO 45003 relate to technology adoption? ISO 45003:2021 provides guidance for managing psychosocial risk within an OH&S management system. Its value here is that it keeps work design, leadership and worker experience inside the assessment.

What is the first question an EHS manager should ask? Ask, “What problem are we solving, and what will people do differently when the tool works?” If the answer is vague, procurement is ahead of diagnosis.

Where can I hear the full discussion? Listen to the full conversation with Cam Stevens on Headline Podcast.

Digital change is already changing how people are monitored, scheduled and asked to decide. Leaders who wait until a tool is embedded before assessing its psychosocial effects will be reviewing a trust problem after it has become operational.

Cam Stevens's Episode 15 gives safety leaders a disciplined starting point. Define the problem, test the signal, protect worker voice and refuse technology that adds data without improving the decision. Listen to the full conversation.