How Cam Stevens Thinks About Technology-Driven Psychosocial Risk at Work

Episode 15 of the Headline Podcast, published on March 18, 2026, features Cam Stevens, safety technologist and CEO of PKG.

His argument changes how leaders should think about technology-driven psychosocial risk because the wrong tool can add pressure, surveillance, and confusion while appearing to improve control.

Technology should begin with a defined problem

Technology is worth considering only after leaders can describe the work problem, the affected people, the decision that needs better evidence, and the control that may change as a result. Stevens rejects technology selection as a catalogue exercise because a device or platform cannot decide whether the organization is solving exposure, coordination, fatigue, competence, or trust. The first question is therefore not which product to buy. It is what unsafe condition the organization is prepared to change.

That sequence is easy to reverse. A new sensor, dashboard, wearable, or artificial-intelligence tool creates visible activity, so the organization begins measuring adoption before it has agreed on the risk. Procurement then asks whether a system can integrate, while operations asks whether it can run at scale. Neither question establishes whether the tool will make a safety-critical decision clearer.

Stevens stated the principle directly on the episode: “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 places responsibility before novelty.

Adoption is not the same as control improvement

A technology adoption is a control improvement only when it changes a decision, strengthens a barrier, or gives a person earlier and better information about a changing condition. A completed installation, a high login rate, or a full dashboard does not prove that the hazard is better managed. Leaders should trace the path from signal to decision and verify whether the decision changes the work.

Consider a platform that flags a rising exposure. If nobody owns the review, the signal becomes another notification. If a supervisor sees it but lacks authority to pause the task, the system produces awareness without action. If workers receive alerts that do not distinguish urgent threats from routine variation, they learn to filter the tool rather than trust it.

Three questions expose the gap. Who receives the signal? What decision must follow? Which authority can change the plan within the required time? A tool that cannot answer all three may still support exploration, but it should not be described as a reliable safety control.

Prediction claims need a visible evidence chain

A prediction is credible when leaders can explain what the system observes, how the observation becomes a risk signal, what uncertainty remains, and which human decision follows. Prediction language becomes dangerous when a probability score is presented as a fact or when a model hides the conditions that produced its output. The responsible test is whether its evidence can be challenged by people who understand the work.

The technology conversation often moves faster than the evidence conversation. A vendor may describe prevention, anticipation, or intelligence while the customer has not defined the baseline, the false-alert tolerance, or the consequence of a missed signal. Without those boundaries, a score can influence work allocation, individual scrutiny, or escalation without a stable explanation.

The review should separate four layers, which are often collapsed into one label. The first is the raw observation. The second is the interpretation applied to that observation. The third is the confidence and limitation attached to the interpretation. The fourth is the operational response. When a leader can inspect all four, technology remains a support for judgment rather than a substitute for it.

That distinction protects workers from being treated as the risk itself. A pattern may indicate that a task, shift, interface, or staffing model needs attention. It does not automatically show that an individual is careless, resistant, or less capable.

Voice technology can reveal work, but it can also change work

Voice technology can become a valuable source of operational evidence when workers use it to report conditions, describe friction, or surface changes that conventional systems miss. It can also create psychosocial risk when people believe every word is being scored, stored, or used against them. Leaders must define purpose, access, retention, and response before asking workers to provide voice data.

Stevens described voice technology as a data source, not as a magic route to truth. Speech carries context, emotion, language differences, and power relationships. A worker who speaks in a second language may be interpreted differently from a worker who uses the vocabulary expected by the system. A contractor may report less freely if access to future work appears connected to the record.

The safest deployment separates reporting from disciplinary surveillance. Workers should know what is collected, why it is collected, who can access it, how long it remains available, and what happens after a concern is raised. Those rules should be visible before launch, not explained after a dispute.

Leaders should test the response loop with one hypothetical report, one anonymous report, and one urgent report before expanding the system. If none reaches a named owner, the organization has purchased listening without creating a reliable response.

Technology can move psychosocial risk into the interface

Technology-driven psychosocial risk appears when digital systems increase demand, reduce control, blur accountability, or make work harder to understand. The risk may sit in alert volume, constant availability, opaque scoring, fragmented instructions, or the fear that a mistake will be permanently recorded. A system review must examine how the interface changes attention, autonomy, communication, and recovery.

Stevens warned that the changing risk profile will be “overwhelmingly psychosocial, driven by technology usage in our organizations.” The point is not that technology is inherently harmful. It is that technology changes the conditions in which people prioritize, communicate, learn, and recover.

The World Health Organization describes working conditions as a factor that can protect or harm mental health, while ISO 45003:2021 gives organizations guidance for managing psychosocial risks within an occupational health and safety system. Those sources support a practical conclusion. The digital layer belongs in the risk assessment when it changes workload, role clarity, supervision, or the ability to stop and ask for help.

In practice, a useful review looks at interruption frequency, duplicate entry, after-hours contact, unclear ownership, alert escalation, and the time needed to correct bad data. These are work-design questions, not requests for employees to become more resilient.

Compare the tool promise with the operating reality

Leaders should compare what a technology promises with what the operating system can support. The comparison needs to include the worker experience, not only the technical specification. If the organization cannot resource the review, protect the reporter, maintain the equipment, or act on the signal, the tool may increase exposure while its dashboard suggests progress.

The table is deliberately operational. It asks leaders to inspect the point where a product meets the work, because that is where psychosocial risk becomes visible. A system may be technically accurate and still be unsafe to use if it adds unresolved demand or transfers responsibility without authority.

Technology should sometimes be refused

A mature technology decision includes a defensible reason not to deploy. Stevens said, “We can elevate the human experience with technology, but there are certainly times when technology should absolutely not be used.” Refusal is appropriate when the problem is undefined, the evidence cannot be explained, the privacy boundary is unacceptable, or the organization lacks the authority and capacity to act on the output.

That decision protects more than budget. It prevents workers from carrying the psychological cost of a system that watches without helping, predicts without explaining, or records without resolving. It also preserves trust, which is difficult to restore after people conclude that participation creates personal exposure.

A leader can document the refusal in four lines. State the unresolved problem, identify the harm the tool could add, name the missing capability, and define the evidence required to reconsider the decision. This turns “not yet” into governance rather than resistance to innovation.

Recommendation

Before adopting a safety technology, run a 30-day problem-first review with one work group, one named decision owner, and one explicit control objective. Record the current condition, the signal the tool would produce, the response authority, the worker protections, and the evidence that would justify expansion. If the review cannot show how the tool changes work without adding hidden demand or fear, stop the procurement and redesign the problem statement.

Start with a small decision, not a large rollout. Ask the people who perform the work what the system might miss, what it might misread, and what would make them stop contributing information. Then compare those answers with the vendor's claims and the organization's actual response capacity.

Stevens's central contribution is a discipline of restraint. Technology is not the strategy. It is one possible instrument in a strategy that must still define the risk, protect the person, assign the decision, and verify the change.

Listen to the full conversation with Cam Stevens on the Headline Podcast, where the discussion examines technology adoption, prediction, voice data, and the psychosocial risk created by digital work.