When artificial intelligence works, who disappears?

We are told it is “restructuring”. The word is clean, cold, almost reassuring. It gives the impression that a company is simply putting its files in order, modernising its procedures, adjusting its costs.

The point is not to block artificial intelligence. The point is to refuse to let it advance in disguise, replacing human work under the comfortable vocabulary of “restructuring”.

But behind that word there is often a less elegant reality: women and men replaced by automated systems. Not always, not everywhere, but often enough that we should stop pretending not to understand.

When a company shrinks its customer service after installing conversational agents, is that still merely restructuring? When administrative posts vanish because software sorts the requests, drafts the replies and prepares the reports, is that still only a question of “efficiency”? When jobs in writing, translation, design, marketing, support or analysis are absorbed by AI tools, must we keep talking as though nothing had changed?

Artificial intelligence no longer merely assists human work: in some sectors, it is starting to replace it. And the problem is not that technology advances. The problem is that workers are pushed back without society having truly debated the direction.

Modernity must not become a polite word for layoffs

The signals are multiplying everywhere — in tech, customer support, financial services, communications, the creative industries, administrative functions. Companies are cutting their headcount while investing massively in automation and AI.

Not every layoff is caused by AI, of course: there are the post-pandemic over-hires, shareholder pressure, economic cycles, and those executives who suddenly discover “financial discipline” after spending like children in a toy shop. But it would be dishonest to deny that AI has become one of the great instruments for cutting and reorganising work. Sometimes it genuinely replaces. Sometimes it serves as an excuse. Sometimes it does both. That is precisely why we need transparency.

The world knows the shock is coming

To grasp what is at stake, a detour through horses is in order. For millennia, every new tool made the horse more valuable: better ploughs, better carts, better roads — and therefore ever more horses. The idea that “work never disappears, it merely shifts” held true for the horse for centuries… until the combustion engine. At that point, horses did not retrain into more skilled equine work. They simply became unemployable. The economist Daniel Susskind, in A World Without Work (2020), draws a warning from this: the mistake is not believing there will be no more work — there will — but assuming that this new work will necessarily fall to us, to humans. No economic law guarantees that a freshly created job will carry your name rather than that of a piece of software.

The International Labour Organization, in its May 2025 report (Working Paper 140), estimates that one job in four worldwide is exposed to generative AI. The proportion climbs to 34% in high-income countries. And the exposure is not gender-neutral: worldwide, 4.7% of women’s jobs fall into the highest-exposure category, against 2.4% for men; in wealthy countries the gap widens to 9.6% versus 3.5%. Clerical, office, communications and customer-service jobs — the ones that employ women most heavily — are the most exposed.

We will be told that the main effect will not be the disappearance of jobs, but their transformation. Very well. Except that transforming a job can also mean: cutting half the tasks, halving the team, asking those who remain to supervise machines, and calling that “productivity”.

Who decides? Who benefits? Who loses their job? Who keeps the value created? Who is retrained? Who is left behind? These are the serious questions — and they are exactly the ones avoided when the conversation is confined to “innovation”.

But there is a decisive difference between us and horses — and that is where everything is settled. The horse did not vote. It did not consume. It could neither demand rules, nor refuse a product, nor choose its leaders. The worker, by contrast, is also a citizen, a consumer and a voter. What was a fate for the horse remains, for us, a collective decision. Technology advances; the way we let it reshape work, however, depends on political choices we still have the power to make. That is precisely why transparency is not a technical detail, but the first of those choices.

Some countries already regulate the use of AI in recruitment and the management of workers; others are betting on continuous training. That is necessary, but it is not sufficient. You do not protect workers merely by telling them “adapt”. You must also tell companies: “take responsibility”.

There is the easy objection: that it would be enough to tax the companies that automate more heavily. Tempting, but inadequate. One more tax keeps no one in their job; it tells no worker that their dismissal was due to a machine; it fills a budget, not an employment contract. And in a country where citizens already struggle to follow the trail of public money, creating a new revenue stream without transparency simply means enlarging an opaque pot. Transparency must come first; taxation, perhaps, afterwards — but never the other way round. You do not cure an illness by hiding its cause.

Our proposal: a label for the consumer

Today, when a citizen buys a product or uses a service — an article, an advertisement, a translation, a design, a customer-service interaction, a preliminary legal or medical opinion — they almost never know whether they are dealing with a human, a human team assisted by AI, or a machine with a human somewhere at the end of the chain to click “approve”.

We demand labels for fair trade, for organic, local, recycled, halal, vegan goods. We want to know the origin, the composition, the environmental impact. Why would ethics stop at the raw material, and not extend to human work?

This is not about banning AI — that would be absurd. It can help doctors, teachers, researchers, public servants, small businesses. But innovation must not become a curtain behind which workers are erased. We therefore propose a simple, legible, enforceable transparency obligation: products and services placed on the market should display one of the three following labels.

This label does not say “do not use AI”: it says “tell us when you use it”. It does not judge the technology; it gives the citizen back the right to choose. Some will prefer a cheaper automated service; others will want to support companies that keep human jobs; others still will refuse to have an artistic creation, a sensitive piece of advice or an essential service handed to a machine. But to choose, one must first know. Today, that choice does not exist — and, as so often, transparency suddenly becomes very complicated when it risks costing something to those who profit from opacity.

And let no one tell us the idea is unrealistic: Europe is already making it mandatory. Article 50 of the European AI Regulation — the AI Act — comes into application on 2 August 2026. It requires that any content generated by artificial intelligence be clearly flagged, and its code of practice draws precisely the distinction between content that is “fully AI-generated” and content that is merely “AI-assisted” — exactly the distinction we propose here. The real question, then, is not whether such transparency is possible, but why Mauritius would wait years before granting its citizens a right that Europe is already enshrining.

And this concerns us more directly than it might seem. The bulk of our offshore sector — BPO, service centres, digital outsourcing — works for European clients; that is precisely why we train a French- and English-speaking workforce. Yet from August 2026, those European clients will be subject to the AI transparency obligation. In other words, Mauritian companies producing for Europe will have to comply with this labelling anyway. The real question is therefore not whether our companies can do it — they will, for their foreign clients — but whether Mauritian citizens and workers will be entitled, at home, to the same transparency our economy already offers Europe. Refusing to legislate spares our companies no effort; it merely guarantees that Mauritians remain the last to be kept in the dark.

Concretely, how?

The principle: self-declaration by the company, random checks by the consumer-protection authority, graduated sanctions (warning, fine, publication of the breach). No new bureaucracy; simply a mandatory field added to the legal notices already required.

The scope: begin with the pilot sectors where generative AI is already heavily used and where human responsibility remains essential — media and advertising, preliminary legal or medical advice, retail financial services, online training. Extend afterwards, sector by sector, on the basis of accumulated experience. No big bang; a gradual, legible roll-out.

And what about Mauritius?

In Mauritius, we rarely hear of layoffs “caused by AI”, in a service economy where BPO, finance, communications, accounting and administration nonetheless employ tens of thousands of people. The explanation is not that it does not happen. The explanation is that the law requires no one to say so.

The Workers’ Rights Act 2019, in its Part VI and notably its Section 72A — Reduction of workforce in certain enterprises in the services sector — allows any employer to cut its headcount in the name of “restructuring”. The 2021 amendment through the Finance (Miscellaneous Provisions) Act even lowered the threshold: the company need no longer be in imminent financial peril; it is enough that the restructuring allow it to “better manage its financial position and avoid financial perils”.

And this is where the real hole appears. You go looking for the form an employer fills in to notify such a layoff to the Redundancy Board… and there is none. No standardised form, no box, no dedicated field. The employer must, to be sure, justify the decision — over-indebtedness, non-viability, reorganisation — but no standardised document captures that justification in a structured way, and nowhere is the employer asked: were automation or artificial intelligence a substantial factor in this decision?

Nowhere. So nowhere is it written. So nowhere is it known. In Mauritius, officially, there is no such thing as a layoff “caused by AI” — not because it is rare, but because the law makes no one ask the question.

This is the real hole in Mauritian transparency. And it calls for exactly the same remedy as the label on products, one step upstream: if we ask a company to state honestly how its service is produced, we can also ask it to state honestly why it is cutting its team. Since no form exists, let us create the one that is missing: a mandatory declaration, submitted to the Redundancy Board under Section 72A, with a single question — “Was automation or generative AI a substantial factor?” Three answers: yes, partly, no. After-the-fact audit in case of doubt. It blocks nothing, penalises nothing — it simply makes visible a reality whose scale the country, today, has no right to know.

Young people will be particularly exposed if the entry-level posts, the ones that let you learn a trade, are the first to be automated. What becomes of a generation told “you must have experience”, but whose first jobs have been handed to machines — with, for sole answer, a training course on “agility in the face of change”?

If AI transforms work, the law must transform transparency. The citizen must know what they are buying and what they are endorsing — and the worker must know what is replacing them.

Progress must not advance in disguise

A restructuring can be necessary, an automation useful, an innovation beneficial. But a society that replaces humans without saying so, that transforms work without debate, that lets citizens consume without knowing, is not a modern society. It is a society advancing blindfolded and calling it vision.

At En Avant Moris, we believe technology must remain in the service of people — not the other way round. AI must be a tool: not an excuse, not a mask, not an elegant way of making workers invisible. Whether a company produces by AI, with AI or by humans, let it have the courage to say so; if it cuts its team because a machine now does the work, let it have the decency to admit it. What we ask is simple: that the citizen know, that the worker know, and that progress stop hiding behind the word “restructuring”.

* That is precisely what we are defending: not hiding AI, not demonising it, but stating clearly when it is involved. Because transparency must not begin with others. It must begin with us.