Research
AI job exposure is not a headcount forecast
Use labor-market research to plan skills and tasks without turning potential exposure into a guaranteed employment outcome.
Zocri · 2 min read · Sources checked
Preserve the meaning of exposure
The IMF's January 2026 discussion says nearly 40% of global jobs are exposed to AI-driven change and examines demand for new skills. Exposure is not a prediction that the same share of jobs will disappear. Tasks can change through assistance, automation or new responsibilities.
For a small employer, a role is usually a bundle of tasks. The person answering inquiries may also recognize a regular customer, resolve an unusual complaint and coordinate a delivery. A tool that drafts an answer changes one part of that bundle. It does not establish what should happen to the whole role.
Map work before changing staffing
Write down the tasks a team completes and the decisions each requires. Separate drafting, checking, communicating and resolving exceptions. Ask employees which information is missing most often and where a mistake is expensive to repair. This makes a pilot more concrete than a broad instruction to become AI-first.
A hypothetical bookings coordinator might spend less time transcribing requests if customers submit structured details. That could create time to confirm complex appointments. It could also introduce a new queue of incomplete forms. Both possibilities need observation; neither should be assumed from an economic headline.
Teach judgment alongside tool use
A useful training exercise includes a plausible wrong answer, a stale record and an ambiguous request. Staff should know when to correct the output, ask a customer for clarification or stop the workflow. Give them a clear route to report repeated problems without having to work around them silently.
Track the workload that moves to reviewers. If experienced staff must inspect every output in detail, their time is part of the cost. Training success means people can complete and recover the real task, rather than simply demonstrating that they can enter a prompt.
Measure what changed for people
Compare the task before and after the pilot using the same definitions. Record completed requests, correction time, unresolved cases and the work staff can now take on. Discuss the findings with the people doing the work, including new starters who may encounter different difficulties.
Use employment research as context for planning capabilities. Your business decision needs evidence about your own tasks and service obligations. A careful task-level trial provides a better basis for training and process changes than multiplying an exposure percentage by the number of people on your payroll.