People and Society lab: how AI is changing people-centred tasks

Key facts

• In the ILO's 2025 index, secondary teachers (0.30), social workers and counsellors (0.28) and nurses (0.25) are rated "not exposed" to generative AI.^[1]
• Contact centre clerks (0.58) and translators (0.59) are in the ILO's second-highest exposure group.^[1]
• Employers in the WEF's 2025 survey expected teachers, social workers and counsellors, nurses and personal care aides to be among the largest-growing jobs to 2030.^[2]
• Half of those employers (50%) counted empathy and active listening as a core skill.^[2]
• Malaysia's Personal Data Protection Act 2010 regulates how personal data is handled in commercial transactions.^[3]

People and Society is about work where the main material is other people: teaching, caring, advising, listening, explaining, organising. This brief looks at those jobs as bundles of tasks and asks how generative AI (tools that write, summarise, translate or talk from a prompt) is changing each one. Every forward-looking line here is someone's estimate, and the estimates could be wrong.

How to read the evidence

• ILO exposure scores. The International Labour Organization's May 2025 index scores how much of each occupation's current tasks generative AI could potentially do, from 0 to 1.^[1] It is a global estimate built from a Polish task list, a worker survey, expert review and an AI model, not Malaysian data.^[1] It found one in four workers worldwide in a job with some exposure, and women more often than men in the most exposed group (4.7% against 2.4% of employment).^[1]
• WEF projections. The World Economic Forum's Future of Jobs Report 2025 asked over 1,000 employers about 2030.^[2] Employers in Malaysia expected 41% of workers' core skills to change in that time, close to the global 39%.^[2] These are expectations, not measurements.
• Anthropic Economic Index. Anthropic, which makes the Claude AI assistant, studies anonymised Claude conversations. In 2025 it found very low AI use in jobs needing a lot of manual dexterity.^[4]
• O*NET. The US occupational database lists the tasks in each job, a useful starting point for your own task list.^[5]

Example tasks and how AI is changing them

"Not exposed" does not mean "safe"

The ILO's lowest groups mean most of today's tasks score low, not that nothing will change. High exposure does not mean a job disappears either: that depends on how employers adopt the tools and whether workers get to learn them.^[1] Care, teaching and counselling still involve tasks AI is changing, such as paperwork and preparing materials. Avoid deciding a career is "safe" or "doomed" from one table.

Consent, privacy and who gets a say

• Never paste people's details into an AI tool. That means real names, IC numbers, addresses, health details or case notes. In the First Move task trial every record is fictional for this reason.
• The law. Malaysia's Personal Data Protection Act 2010 regulates the processing of personal data in commercial transactions, and the Personal Data Protection Department (JPDP) administers it.^[3] Check the current rules on the JPDP site before you handle real data at work.^[10]
• The guidelines. Malaysia's National Guidelines on AI Governance and Ethics (MOSTI, September 2024) are voluntary and set seven principles, including fairness, inclusivity, privacy and data security, and accountability.^[7]
• Ask first. Ask before you record, photograph or quote someone, and tell them if AI will process what they said.
• Ask who is left out. A digital fix can exclude people without phones, data or the language it uses. Ask them, not only the people who are easy to reach.

Costing and pitching

• Costing. People services cost staff time. If AI saves an hour of paperwork, ask who gets that hour: more time with clients, or fewer staff? Estimate the time each step takes before and after.
• Pitching. Funders and managers want to know who benefits, who might be left out, and what it costs to run each week.

In the taster

In the People and Society taster you teach a four-minute lesson. You ask AI for a lesson plan, then test it by following its steps exactly, the way a pupil would. Then you check whether your pupils understood, not just nodded. That is how teachers and trainers use AI: a draft to test on real people.

For facilitators

• This lab pairs with the community service centre task trial: missed appointments, mixed-language messages and visitors with different digital access. Ask learners to check one AI translation and one AI number against the supplied information, and to say who a digital fix might exclude.
• Use the table as a starting list for the Job X-ray and have the practitioner correct it first. Learners support one label with a source and challenge one of their own assumptions.
• Anthropic's Econ Scenario Explorer (September 2026) is a simplified US model of 2030 that changes with your assumptions.^[11] Use it to show that assumptions matter, not as a Malaysian forecast.

What this means for you

• The human core of these jobs is the least exposed. Trust, care, judgement and knowing a particular person are hard for AI to take over, on the ILO's measure.
• The paperwork and first drafts around them are changing. Expect to use AI for notes, materials and translations, and to be the one who checks them.
• Routine front-line service is the most exposed. Customer service and translation drafts are changing fastest, so build the skills for the hard cases.
• Privacy is part of the job. Handling people's data carefully is a skill employers in this lab will expect from day one.

Your next step

1. Pick one task from the table that you could see yourself doing. On Day 2, ask the practitioner which part of it AI already helps with, and which part they would never hand over.
2. Read how AI may change your career area: social impact, health and science, business and enterprise or global opportunities.