Higher productivity is an easy ambition to applaud until the productivity gain is your job. Producing more with less effort has generally meant greater prosperity, with cheaper goods, better wages and new work replacing the old. Artificial intelligence could accelerate the first part of that process without delivering the rest quickly enough for the people whose livelihoods depend on it.

A company that produces the same work with half its staff has improved its output per employee. Whether the country has become better off depends on what happens to the other half. Repeat that calculation across enough businesses and a succession of sensible commercial decisions could leave the economy struggling to sustain the spending that makes them profitable.

The reassurance that technology creates jobs is perfectly true and entirely insufficient. It says nothing about whether those jobs will arrive in sufficient numbers, pay enough or be accessible to someone whose previous expertise has just become considerably less expensive to obtain.

This does not require machines to become better than humans at everything, although assuming today’s limitations are permanent would be a peculiar bet to place on a technology developing this quickly. Accountants, lawyers and designers can remain necessary while considerably fewer of them are needed.

The division of labour has been preparing the ground for centuries. Businesses broke production into specialised tasks, refined those tasks into procedures and employed people to supply the judgement and flexibility machines lacked. Much professional work now involves gathering information, applying rules, interpreting results and passing them along. Breaking a job into steps does not make every step safely automatable, but an impressive title does not exempt its holder from the exercise.

AI needs to perform enough of those tasks for one experienced person to supervise work previously requiring a team. In creative work, it need only produce something customers find adequate at a price a person cannot match. A designer may win the argument about whether machines possess imagination and still lose the commission.

For tasks AI can already perform reliably, protection often lies in the difficulty of putting it to work. A company-wide ChatGPT subscription does not reorganise a business: integration requires investment, suitable data, testing and clear responsibility. Professional warrants and legal obligations impose further constraints. These are substantial barriers, but they offer a different kind of reassurance from the claim that human beings possess some commercially indispensable magic.

An IMF study published in 2025 identified roughly 30% of Malta’s employment as particularly exposed to displacement: occupations where AI capabilities overlap substantially with the work, with fewer features favouring continued human involvement. Clerical and administrative roles featured prominently. This identifies vulnerability, rather than forecasting that three in ten jobs will disappear.

That should trouble a country which has spent years encouraging young people to obtain degrees and supporting them with stipends. Education remains valuable, but firms could continue needing experienced professionals while recruiting fewer juniors to become them. The qualification might survive rather better than the career ladder.

The “I should have become a plumber” joke becomes less amusing when a 14-year-old asks which subjects to choose. Some skilled trades are harder to automate than work traditionally associated with qualifications and security, yet nobody can confidently map today’s opportunities onto the labour market that child will enter.

The future could turn out considerably better than this suggests. Cheaper tools could help small firms compete, attract new customers and create work. But whether those possibilities translate into broadly shared prosperity depends on decisions about ownership, institutions and bargaining power, as well as technical capability.

Competition among the wealthy and powerful did not begin in Silicon Valley. States and laws have restrained it while also protecting accumulated advantages, and markets have repeatedly rewarded businesses whose costs were borne by ecosystems, communities and, increasingly, the public’s capacity to distinguish information from manipulation. It would be optimistic to assume the distribution of AI’s gains will somehow look after itself.

AI could make it easier for wealth to access skills while making it harder for people with skills to earn money. Businesses collectively need customers with purchasing power, but no individual employer is obliged to preserve a payroll for the benefit of somebody else’s customers.

In Malta, weaker earnings would affect shops, restaurants and housing. Departing foreign workers could reduce demand while keeping unemployment figures deceptively reassuring. Lower rents would help tenants, but vacancies could strain indebted landlords, weaken construction and spread the effects to suppliers and lenders.

Malta’s economic structure could provide a temporary cushion. Tourism requires physical services, while internationally active firms might maintain sales and profits despite reducing local payrolls. Government could retain revenue with which to support households, although tourists have jobs abroad, clients can cut spending and companies can relocate. Healthy receipts could buy time or merely encourage us to waste it.

Where revenue holds up, Malta’s much-maligned public payroll might acquire an economic justification its critics had overlooked. Government could use AI to improve services while retraining and redeploying staff, turning productivity gains into shorter queues rather than fewer salaries. This requires useful work and competent management; creating another office where nobody answers the telephone would miss the point.

Temporary earnings top-ups could help people taking lower-paid jobs, while paid training and placements could preserve routes into professions. Broader, lasting disruption would require serious consideration of an adequate income floor. Each option needs scrutiny: top-ups can subsidise poor wages, courses can enrich providers without helping participants, and permanent support requires permanent funding.

Collecting more from automation’s profits may become necessary, but Malta cannot assume internationally mobile firms will stay to pay. Government needs to weigh revenue against relocation risks, pursue international coordination and monitor recruitment, hours and earnings before unemployment makes the problem obvious.

These choices give spending discipline a purpose beyond the annual performance of fiscal seriousness. Social support has delivered real benefits, but money lost to corruption, waste and political theatre reduces the resources available for a transition that may demand considerably more public intervention.

Much of its pace will be determined by technology companies and governments competing for supremacy. They will not wait for Malta to redesign its education system, although government here will inherit consequences it had little power to prevent.

Work provides company, purpose and recognition as well as income, so replacing a salary would address only part of the loss. Greater productivity could give people more freedom to care, create and participate, provided they retain security and a meaningful place in society.

An economy requiring less human effort ought to be an achievement. Whether people experience it as liberation or redundancy will depend on who controls the gains and what governments do while they still have room to act.