“This job is exposed to AI” is often read as “this job will disappear”. The first statement concerns what a technology might affect. The second predicts what employers, workers and customers will do.
That gap matters when interpreting reports about the future of work.
A job is a bundle of tasks
Someone coordinating an office may write emails, resolve exceptions, maintain supplier relationships and notice when a colleague needs help.
AI might assist with parts of the writing and record handling. It does not follow that the entire role can be removed or that the remaining work stays the same.
The International Labour Organization’s 2025 occupational exposure study evaluates tasks and distinguishes exposure from inevitable job loss. Its findings point to transformation as a likely outcome for many affected occupations.
Ask what the evidence measures
A capability test may show that a model completes a task under defined conditions. An employer survey may show intentions. A labour-market study may examine observed changes.
Those are different kinds of evidence. A projection should not be presented as an event that has already happened.
Global results also need care when applied to Australia, where industries, job design and workplace arrangements differ.
Discuss changes with the people doing the work
For a hypothetical small professional-services team, map a normal week. Identify drafting, checking, customer interaction and exception handling separately.
Ask where an assistant might help, where additional review would be created and what skills need to remain available. Involve the staff who understand the difficult cases.
If drafting becomes faster, work may shift towards review. That creates a new question about workload and judgement, rather than proving that total effort has fallen.
Follow one request through the revised job
Consider a hypothetical service coordinator handling a late-delivery enquiry. An assistant drafts a reply from tracking details; the coordinator discovers that the promised date conflicts with a supplier message.
Drafting has become easier, but the unresolved work is checking the evidence, contacting the supplier and deciding what to tell the customer. If the team measures only how quickly the draft appeared, it misses the bottleneck.
A better trial records the complete request: preparation, drafting, correction, escalation and resolution. That shows whether effort moved, fell or increased.
Preserve a route for learning
Junior tasks often help people build expertise. If AI takes over first drafts, teams need another way for less experienced staff to practise and receive feedback.
Judge a pilot on service quality, errors, workload and learning as well as speed. Avoid using activity counts as a substitute for understanding the work.
The useful conversation is specific: which tasks change, who remains accountable, and how people gain the skills to do the revised job well.