When Pay Parts from Output: What Does the Evidence Say About Automation?

We grew up on a simple rule: whoever works, eats. On it rest taxes and pensions and insurance, and by it a person's share of wealth was reckoned. What becomes of the rule if a larger part of production comes from machines and algorithms? A serious question, but its answer is more exacting than the headlines.

What has actually happened?

Automation is no deferred promise; it has been affecting wages for decades. A study by Acemoglu and Restrepo attributes between 50% and 70% of the change in the US wage structure, from 1980 to 2016, to the falling relative wages of groups specialised in routine tasks in industries that automated quickly. The study concerns automation in general, not generative AI, and the United States in particular, so its figures should not be carried over to other countries without care. Yet it says that automation can change who gains and who loses even when it does not abolish jobs.

What do the projections estimate?

The World Economic Forum's The Future of Jobs Report 2025, which surveyed more than a thousand employers, expects structural change to touch about 22% of today's jobs between 2025 and 2030: about 170 million jobs created and about 92 million displaced, a net gain of around 78 million. About 39% of workers' current skills are expected to change or become obsolete. Notably, 40% of employers expect to reduce their workforce where AI can automate tasks, while 85% plan to prioritise upskilling.

These remain survey estimates, not settled results. Earlier estimates varied widely, and the article "Artificial Intelligence a Decade from Now: Three Possible Paths" sets out that divergence between methods.

As for exposure, the International Monetary Fund estimates that about 40% of jobs worldwide are exposed to AI, about 60% in advanced economies and about 26% in low-income countries. Exposure does not mean disappearance: half of the exposed jobs in advanced economies may see productivity rise through complementarity with the tool, while the other half may lose their core tasks. The Fund judges that AI is likely to worsen overall inequality in most scenarios, because higher earners may capture a larger share of the gains.

Taxes that lean towards the machine

Studies note that tax systems are not neutral. Acemoglu, Manera and Restrepo concluded in a 2020 paper that the US tax system is biased against labour and in favour of capital, and has encouraged excessive levels of automation. They estimated that moderate reforms could raise employment by between 1.14% and 1.96%. This is research on the American system specifically, but it warns that tax incentives can tempt us to dispense with people before that is the better course.

Is it enough to give people money?

A basic income is often proposed as the remedy. The strongest field experiment so far, led by Eva Vivalt and colleagues, gave a thousand low-income people in two American states a thousand dollars a month for three years, and compared them with two thousand people who received 50 dollars a month. Labour-force participation fell by 4.2 percentage points, hours worked fell by about one to two a week, leisure time grew most, and no improvement in job quality was observed.

This result does not refute the idea of solidarity, but it is a reminder that money alone creates no skill and opens no door to better work. It is an experiment in one context, and does not measure the effect of money combined with training and opportunity. Our own heritage of solidarity, such as zakat and waqf, has a place in this debate that deserves a study of its own. On the meaning and worth of work once machines master tasks, see the article "A Job Is One Thing, Work Another: What Remains When Machines Master Tasks?".

What can be done now?

• Measure exposure locally: which tasks in our sector are affected? Global figures do not replace the local picture.
• Training linked to the job: invest in skills that complement the tool, not in detached certificates.
• Review incentives: our policies should not reward dispensing with people where their presence is needed.
• Safety nets that preserve dignity: and keep the door to work open.

Whoever relies on God while taking the means need not claim certainty about what only He knows; it is enough to prepare with knowledge and caution.