Safety Conversations: 7 Scripts That Change Behavior

A safety conversation is not a nicer way to say "correct the worker." It is a leadership tool for discovering why a risky action made sense to the person doing the work, which is the only way to change behavior without driving the real reasons underground.

This article is written for supervisors, EHS managers, and operational leaders who need field dialogue that changes exposure. The thesis is direct: most safety conversations fail because they start too late, focus too narrowly on the person, and ignore the work pressure that shaped the behavior.

Why most safety conversations do not change behavior

Many supervisors were taught to observe, correct, record, and move on. That sequence looks disciplined, although it often misses the decision pathway behind the behavior. A worker who bypasses a barricade, skips a verification step, or accepts a cramped body position may be making a poor choice, but that choice may also reflect layout, time pressure, weak planning, unclear authority, or a control that only works on paper.

As Andreza Araujo argues in Guide to Behavioral Observation: VAMOS A HABLAR?, behavioral observation must be built around dialogue rather than surveillance. The Spanish title means "Shall we talk?", and that phrase matters because behavior changes when the conversation exposes how the worker understood the risk, not when the observer simply marks a card.

Across 25+ years leading EHS at multinationals, Andreza Araujo has seen that field leaders often notice the unsafe act faster than they notice the condition that made the act predictable. The script matters because a careless question can create silence, while a precise question can reveal the weak signal that prevents the next serious event.

1. Start with the observed fact, not the accusation

The first script is useful when the supervisor has seen a risky action but does not yet understand the context. Say, "I saw you reach across the moving area to clear the jam. Walk me through what made that feel like the available option."

The wording is deliberate because it names the fact without judging intent. It also asks the worker to reconstruct the decision, which helps the supervisor see whether the exposure came from habit, pressure, tool design, production urgency, or a missing isolation step.

This approach connects directly with risk perception in routine work. When the task is familiar, the worker may see the hazard and still discount it because nothing happened the last fifty times. The conversation has to surface that pattern before correction will stick.

2. Ask what changed in the task today

The second script is for jobs that look routine from a distance. Say, "What is different about this task today compared with the last time it was done safely?"

This question prevents the supervisor from assuming that yesterday's method still controls today's exposure. Staffing, weather, equipment condition, materials, schedule pressure, fatigue, contractor mix, or a minor layout change can turn a routine task into a different risk profile.

In a 220-employee packaging operation, for example, a palletizing task may look unchanged while the actual risk has shifted because one lift table is out of service and the crew is rotating through awkward reaches. A generic reminder about posture would miss the real control failure.

Use the answer to decide whether the task needs a new control, a short pause, or a supervisor escalation. If the worker says nothing changed but the exposure is visible, the supervisor should ask one more question: "Which part of this job could hurt someone today if the timing goes wrong?"

3. Test whether the control is real or symbolic

The third script is for situations where a procedure exists but the field condition suggests the control is weak. Say, "Show me how this control protects you if the job becomes noisy, rushed, or interrupted."

That question moves the conversation from compliance to control effectiveness. A checklist, permit, barricade, spotter, or PPE rule may exist, although the real test is whether it still works when the task becomes messy.

This is where safety conversations should connect with control effectiveness metrics. The supervisor is not only coaching behavior. The supervisor is collecting evidence about whether the operation is relying on memory, courage, or paperwork instead of a working barrier.

In A Ilusao da Conformidade, or The Illusion of Compliance, Andreza Araujo warns that a signed form can create the appearance of discipline while the field remains exposed. A strong safety conversation breaks that illusion because it asks the worker to demonstrate the protection, not only confirm the rule.

4. Separate pressure from personal choice

The fourth script is for moments when the worker clearly chose a shortcut. Say, "What pressure made the shortcut feel reasonable, and what would have made the safe method easier to choose?"

This does not excuse unsafe behavior. It makes accountability more accurate because it separates conscious disregard from a predictable adaptation to pressure. A worker may still need correction, although the organization also needs to know whether its signals are rewarding speed over control.

The same logic appears in production pressure and safety shortcuts. If supervisors only tell workers to slow down while daily targets punish delay, the conversation becomes theater. The worker hears the official message and the real message at the same time.

During Andreza Araujo's PepsiCo South America tenure, where the accident ratio dropped 50% in six months, leadership rhythm mattered because people watched what managers reinforced under pressure. Safety conversations should therefore ask about pressure explicitly, since hidden pressure keeps recreating the same behavior.

5. Use one question to identify normalization of deviance

The fifth script is for repeated small deviations that no longer create concern. Say, "How long has the team been doing it this way, and what changed after people stopped seeing it as unusual?"

Normalization of deviance is dangerous because the risky method becomes the local standard. The supervisor may think the worker is making an individual choice, while the crew experiences the shortcut as the way the job is done here.

James Reason's work on latent failures helps leaders avoid the operator-error trap because the visible act often sits at the end of a longer chain. The question should uncover when the unofficial method appeared, who knew about it, and which weak signal was ignored.

When the answer is "we always do it this way," the supervisor has found cultural data. The next step is not a lecture. The next step is to stop the exposure, understand why the official method lost credibility, and assign an owner to repair the control.

6. Close with ownership and a visible action

The sixth script prevents the conversation from becoming an interesting talk with no operational consequence. Say, "What will change before this task is repeated, who owns it, and when will I verify it?"

Behavioral coaching loses force when workers see the same problem return after every observation. If a worker reports that the tool is wrong, the lifting point is unclear, or the work area forces line-of-fire exposure, the supervisor must close the loop with visible action.

The action can be small, but it must be concrete. Replace a missing guard, change the staging point, revise the pre-task brief, escalate a staffing constraint, or remove a defective tool. A promise to "raise awareness" is not enough because awareness does not remove exposure.

In more than 250 cultural-transformation projects supported by Andreza Araujo's team, a recurring pattern is that trust rises when leaders act on what workers reveal. The conversation teaches culture only when the workforce sees that speaking honestly changes something.

7. Record the learning, not only the correction

The seventh script is for the observation record. After the conversation, write one sentence that captures the learning: "The worker bypassed the marked route because material staging blocked the safe path, and supervision had accepted the detour for two weeks."

That sentence is more valuable than a generic note saying "employee coached on safe access." The first version gives EHS and operations something to fix. The second version protects the appearance of activity while leaving the exposure untouched.

Behavioral observation programs often fail because they measure volume before quality. A plant may collect 500 cards a month and still learn little if the records contain only PPE reminders, attitude labels, or vague coaching notes.

Review a sample of records every month. Look for evidence that observers asked about pressure, task change, control weakness, and ownership. If the records do not show those elements, the supervisor training needs to change before the next campaign asks for more observations.

Field script comparison

Conclusion: the conversation is a control test

A good safety conversation does more than correct behavior. It tests whether the worker sees the risk, whether the control works, whether production pressure is distorting judgment, and whether leadership will act on what the field reveals.

These conversations work better when leaders apply psychological safety boundaries, because the worker should feel safe to speak while the unsafe condition still receives a clear correction.

For organizations that want behavioral observation to become a learning routine rather than a card-counting exercise, Andreza Araujo's books, Safety School, and ACS Global Ventures diagnostics help supervisors connect dialogue, risk perception, and visible field leadership. Safety is about coming home, and that requires conversations that make the real work visible.