An illustrative bottling-plant scenario. Priya Shah, Daniel Brooks and Michael Chen are fictional characters; the events and dialogue are invented, not a client account.
At the morning meeting, Michael Chen, the plant manager, introduces a proposed AI-assisted planning pilot. It would draft the next day's bottling schedule from orders, stock records and expected line capacity.
Priya Shah, the production planner, looks down at her notebook. She spends part of every afternoon bringing those records together. But she also knows which packaging delivery needs a phone call and which changeover needs more time than the standard allows.
"Once I've shown it everything I do, what happens to my job?" she asks.
Michael had prepared to discuss saved hours. He had not prepared an answer about her future role.
Later, Daniel Brooks, the line lead, brings up a different concern. A suggested sequence could look efficient on screen while leaving too little time for a required cleaning changeover.
"If I say the schedule won't work, will that still count?"
The pilot has not started. Yet two questions are already on the floor: Will my work still be valued? And will my judgement still be heard?
A factory's AI discussion can sound very different depending on where you sit.
For an owner, the question may be how to reduce time spent preparing information. For a planner, purchasing coordinator, or administrator, the question may be what that means for the value of their work. For an operator, it may be whether the next system will help them make a sound decision—or make it harder to question one.
Those questions belong in the readiness conversation.
Look closely at the work behind the job title
The ILO's 2025 research identifies clerical occupations as particularly exposed to generative AI. But exposure describes tasks a technology might perform. It does not establish that a whole job will disappear. The ILO expects transformation to be the more likely overall effect. ILO–NASK research.
In a factory, preparing a purchase-order update is only part of purchasing. A person may also know which supplier needs a phone call, which substitution needs approval, and which apparently minor delay will stop tomorrow's production.
The same distinction matters in planning, quality administration, and maintenance coordination. Before assigning work to a tool, make the full task visible—including the exceptions and judgement around it.
That also gives the employee a concrete discussion to join. "We are testing assistance with this task" is more useful than an announcement about becoming an AI-enabled business.
Uncertainty deserves an honest answer
It would be wrong to tell people there is nothing to worry about. Technology can change staffing, responsibility, workload, and the route into a career.
It would also be wrong to treat every exposed task as a disappearing job. Statistics Canada's early employment analysis found growth across AI exposure groups through December 2025, while emphasizing how difficult it is to separate AI's effect from other economic changes. That does not guarantee any individual's future. Statistics Canada analysis.
The useful response is clarity about this workplace: what is being tested, what management has decided, what remains open, and how affected people will be involved. If staffing changes are possible, a training announcement is not an answer to that concern.
Change anxiety can become hesitation to speak up
"Change anxiety" is a useful plain-language description of the uncertainty in this scenario. "Hesitation to speak up" describes the behaviour a leader can actually investigate. Calling it a culture freeze too early can turn a question into a diagnosis.
A quieter meeting does not necessarily mean a team is comfortable with the change.
Research involving 402 employees in South Korea linked AI-related job insecurity with knowledge hiding and lower psychological safety. It was a study across industries, not proof of what will happen in an Ontario factory. It does give leaders a reason to take the question seriously. Original study.
Psychological safety means people feel able to take interpersonal risks, such as admitting uncertainty or challenging an assumption. Its relationship with learning was demonstrated in an earlier study of manufacturing teams. It is different from job security. Manufacturing-team research.
Could uncertainty in the office influence the floor? It is a plausible risk where people depend on each other every day. The research cited here does not establish that chain. A plant leader should ask and observe, rather than assume.
Pay attention when someone stops volunteering an exception, avoids questioning a recommendation, or asks privately what will happen once their knowledge is documented. These are prompts for a conversation, not evidence of a bad attitude. Check workload, the usefulness of the tool, and past management responses too.
Give the team a real part in the pilot
Start with one workflow and bring the affected roles into the discussion early. OECD research associates worker consultation with more positive reported AI outcomes, although it cannot prove consultation caused them. OECD survey.
In the bottling scenario, Michael needs to answer Priya's question about her role separately from Daniel's question about decision authority. Inviting both into a pilot is a useful start; it does not substitute for an honest answer about staffing.
Before starting, answer five practical questions:
- Which task are we trying to improve, and what problem does it cause today?
- Which decisions remain with people, including when they can reject an output?
- What do we know about the effect on roles, workload, and staffing?
- What learning time and support will each affected person receive?
- How will we respond when someone reports an error or concern?
Then make the answers visible in daily work. Give an employee who challenges an output a timely response. Record what was learned. Explain what data the tool collects and how it will be used. If a promise changes, explain that too.
The pilot review should include the quality of the work and the quality of participation. Are people reporting awkward cases? Can they explain the response route? Are unresolved concerns being answered? Fewer questions alone are not a measure of confidence.
AI readiness starts with operational readiness. That includes the people who make operating reality visible: their judgement, their willingness to speak, and their confidence that participation will be treated fairly.
Before the next AI pilot, put this question on the agenda: What might our people be afraid this change means—and have we given them a straight answer?