Activity B kit: the Job X-ray

What this activity is for

Activity B, the Job X-ray, is how learners look at the future of work without pretending anyone can predict a safe career. They take one job, break it into tasks, and label how AI might touch each task. Then they support one label with evidence, challenge one of their own assumptions, and let a practitioner correct them. The result goes into "The future" in the Future Map: a reasoned view of how tasks in an option may change, with the uncertainty written down.

The journey template uses it twice:

• Day 1, "Tasks, not job titles" (30 minutes): a light practice run. Learners list about eight tasks for a job they choose, label them, and see which label moves when the facilitator changes the setting (for example, many customers are elderly, speak different languages or have no smartphone).
• Day 2, the Job X-ray (75 minutes): the full version, starting from a task list a practitioner has checked, with the practitioner correcting learners' labels.

The four labels

Two labels on one task are fine. Every label needs a reason, evidence (seen, read or told) and how sure the learner is: low, medium or high.

Why tasks, not job titles

Jobs are bundles of tasks, and AI tends to change some tasks inside a job long before it changes the whole job. Recent research points the same way, with real uncertainty:

• The International Labour Organization estimated in May 2025 that about one in four workers worldwide (25%) are in an occupation with some exposure to generative AI, rising to 34% in high-income countries, while only 3.3% of global employment falls into the highest-exposure category^[1]. Its authors expect most jobs to be transformed rather than eliminated, because most occupations still need human input^[1]. That is an estimate of exposure, not a count of jobs lost.
• The World Economic Forum's Future of Jobs Report 2025, based on a survey of more than 1,000 large employers across 55 economies, projects about 170 million new roles and 92 million displaced by 2030, a net gain of about 78 million, and employers expect about 39% of current skill sets to change by 2030^[2]. These are employer expectations, not observed outcomes.
• An observed US pattern: Stanford's Digital Economy Lab reported in August 2026 that employment for workers aged 22 to 25 in highly AI-exposed occupations was about 19% below where it would be if it had kept pace with similar workers in less-exposed jobs^[3]. The authors stress these are descriptive patterns, not causal estimates, and that the declines are concentrated where AI is used to automate tasks; where AI complements workers, employment was flat or rising^[3]. It is US payroll data, not Malaysian.

Tell learners plainly: nobody knows exactly how a job will look in 2030. Good thinking says what is projected, what has been observed, by whom, and how sure anyone is.

Different assumptions, different futures

Anthropic's Econ Scenario Explorer (version 1.0, September 2026) lets users set five assumptions, about what AI can do, how widely it is adopted, how much it acts alone, how much it raises productivity, and how long people take to find new jobs, and see the US economy in 2030 that those assumptions imply^[4]. The page calls its model "a stark simplification of a complex reality"^[4]. If you use it, show it on one facilitator screen only, to make the point that changing the assumptions changes the future. Never use its outputs as Malaysian wage or job forecasts.

Day 2 run sheet (75 minutes)

If a practitioner cannot reach every pod, learners compare their labels with the practitioner's own pre-tagged list of ten tasks from their week. The worksheet is in job-x-ray-worksheet.md.

A worked example (draft, to be validated by a practitioner)

example-community-centre-officer.csv breaks the job of a community service centre officer, the role behind the Day 2 task trial, into ten tasks with draft labels. For example, greeting visitors and noticing when someone may be at risk are labelled "needs a physical or social setting"; sending reminders "might be automated"; translating a notice "AI can assist", because AI drafts quickly but a fluent person must check; and checking AI-drafted messages for errors "new work".

This example has not been validated. It was drafted for the classroom. Before learners see it as anything but a draft, a practitioner who does this work corrects the task list and the labels, and you record their role and the date. Until then, present it as "our draft, which a practitioner will challenge".

What the facilitator looks for

The main criteria here are "uses and verifies AI appropriately" (does the learner weigh the AI's counter-argument, or simply accept it?) and "explains a justified next step". Record emerging, demonstrated with support, or demonstrated independently, with an example. Strong work:

• lists tasks a person actually does, not vague qualities such as "communication";
• gives a reason and evidence for each label, and marks guesses;
• uses "how sure" honestly, with some lows;
• changes at least one label after the practitioner's feedback, or explains clearly why not.

Paper route

Print the task list as cards. Learners sort them into four labelled piles on A3 paper and write a reason and how sure they are on each card. They fill in the worksheet's "support one label" and "challenge one assumption" boxes; if AI is unavailable, a partner argues the opposite instead. The practitioner corrects the sheet in pen, the learner explains one label aloud, and the facilitator photographs the sheet once devices work.

What to do next

1. Practitioners: before Day 2, check the task list for your field (about ten tasks) and tag ten tasks from your own week, so learners can compare their X-ray with yours.
2. Facilitators: collect one strong X-ray per pod, with the learner's permission, as a sample for calibrating judgements across pods.
3. Learners: in your Future Map, add one task you want to get better at and one question about it for your next practitioner conversation.