The Algorithm Sets Your Pay, Then Fires You — And It Owes You Absolutely Nothing

## Short-Term Outlook (1–6 Months)

The first verdict arrives in Geneva. The 114th ILC is effectively finalizing the structural architecture of the platform convention between June 1 and 11, 2026, and in this short-term window, I believe the dual structure — a binding Convention paired with a non-binding Recommendation — is likely to survive as a framework. The real fight is over what ends up inside each document. Whether algorithmic transparency and workers' right to contest automated decisions appear in the binding convention text, or get quietly shuffled into the recommendation column, will set the direction of platform labor law for the next five years at minimum.

My base case for this short-term window is what I'd call a "half-win": the convention gets formally adopted, but its most critical transparency provisions end up in the non-binding annex rather than the binding text. The current negotiating dynamics point clearly toward this outcome. The United States, Argentina, and Pakistan are pressing for the weakest possible enforcement mechanisms, and the International Organisation of Employers — representing companies from 150 countries — has publicly complained that the draft is "too long and too prescriptive." On the other side, the EU, Brazil, Mexico, and the International Trade Union Confederation are pushing hard for robust provisions. With that split, the realistic short-term outcome is a diplomatic compromise: strong principles announced in the preamble, weak enforcement in the operative clauses.

There's one important wildcard in this short-term picture, and I think it gets underestimated: the speed and precision of the public reaction. The moment it becomes widely reported that algorithmic transparency has been quietly downgraded to a recommendation, labor unions and digital rights organizations will launch counter-campaigns almost immediately. I expect a very specific, unified demand — "restore algorithmic transparency to the binding text before the final plenary vote" — to become a prominent international talking point within days of that news breaking. Whether that concentrated pressure is enough to flip even a small number of delegations before the final vote is genuinely uncertain, but I wouldn't call it impossible, especially if EU member states hold firm and the coalition coordinating around Human Rights Watch and Privacy International maintains its discipline.

## Mid-Term Outlook (6 Months to 2 Years)

Adoption is only the opening act. ILO conventions carry no legal force until individual countries formally ratify them and transpose them into domestic law. Over the next two years, ratification dynamics will almost certainly polarize sharply, mirroring the split we already see in platform labor regulation more broadly. EU member states will face internal pressure to ratify quickly, propelled by the December 2026 Platform Work Directive implementation deadline. Countries like Mexico, which has already enacted employee classification legislation, are natural early movers with little political cost to ratification.

The EU directive implementation will be the most closely watched test case of the mid-term period. When all member states are legally required to ban algorithmic terminations and mandate human oversight of automated decisions by December 2026, the world will see for the first time, at large scale, whether those requirements can actually be enforced in practice. If EU implementation succeeds and creates measurable improvements in platform workers' conditions, it becomes a powerful demonstration effect for countries still deliberating. If enforcement proves toothless or inconsistent even within the EU's strong regulatory framework, it will hand ammunition to skeptics everywhere who argue that algorithmic regulation is inherently impractical.

Meanwhile, the US and China — home to Uber, DoorDash, DiDi, and Meituan — are likely to delay ratification indefinitely or reject it outright. This matters because regulatory vacuum compounds over time. Uber's commission rate climbed from 32 percent to 42 percent between 2022 and 2024 even as public pressure grew. In a prolonged absence of binding legal constraints, the algorithmic extraction of worker earnings becomes progressively more sophisticated and harder to reverse. I expect some platforms to breach the 50 percent take-rate threshold on specific route categories within the next two years in markets without strong regulation — and once that level becomes normalized, the political effort required to push it back down increases substantially.

A second important mid-term variable is worker organizing capacity. Platform-based unions like Mexico's UNTA, founded in 2020 and already influential in its domestic context, are beginning to spread their model to other developing economies. EU algorithmic transparency requirements will generate litigation as workers and unions use the new legal framework to challenge specific algorithmic decisions in court. I expect the next two years to produce a pattern where courts and unions accomplish through enforcement pressure what regulation alone hasn't yet managed — establishing real accountability for algorithmic management decisions, case by case, jurisdiction by jurisdiction.

## Long-Term Outlook (2 to 5 Years)

Let's start with the number that changes everything structurally: one billion. The World Bank projects that platform workers could exceed 1 billion by 2030. The World Economic Forum suggests the gig economy could represent half of all global employment. The platform economy market is projected to grow from $485 billion in 2025 to $875 billion in 2030 at a compound annual growth rate of 11.1 percent, potentially reaching $1.39 trillion by 2035. My core long-term argument is straightforward: you can politically ignore 435 million gig workers, but you cannot politically ignore 1 billion of them. That constituency, at sufficient scale, generates enough electoral weight, union organizing capacity, and consumer leverage to move legislative bodies that are currently unmovable.

But scale doesn't automatically translate into rights, and the long-term picture forks into three distinct scenarios. In the bull scenario, the EU's platform work directive and Mexico's employee classification legislation become the effective global template. Countries that initially hesitated ratify the ILO convention as domestic political pressure builds with workforce growth. The weak clauses of the ILO framework get strengthened through national legislation, and algorithmic transparency gradually becomes recognized as a foundational labor right. Within five years, the employment presumption principle has become common sense in the majority of jurisdictions accounting for most platform work volume, and platforms are legally required to explain algorithmic decisions that affect workers' livelihoods.

In the base scenario — which I consider most probable — a Rajasthan-style compromise becomes the international template. Just as India's state of Rajasthan introduced a welfare levy of 1 to 2 percent per transaction in 2023, many countries settle for creating welfare funds without touching the underlying employment classification structure. Social insurance coverage rates inch upward. The fundamental architecture — algorithm acting as employer, corporation evading employer obligations — remains intact. This compromise is politically easy to achieve. A small per-transaction levy is something even major platforms can absorb as a cost of doing business, and it provides governments a headline win without requiring a direct confrontation with powerful tech industry lobbying operations. But this approach leaves workers as beneficiaries of a welfare fund rather than citizens with rights, and it sidesteps the central question rather than answering it.

In the bear scenario, this convention follows the path traced by Convention 189. Adopted with optimism, celebrated in press conferences, and then quietly left to gather dust as ratification stalls at a fraction of member states. With the United States and China outside the convention's legal framework, the world's largest platform companies face no binding obligations. The $205 million Prop 22 playbook gets replicated in jurisdiction after jurisdiction — every time a legislature comes close to meaningful platform worker protections, a well-funded campaign emerges to soften, delay, or reverse the legislation. The current baseline — only 35 percent of platform workers with any pension coverage, more than half lacking health insurance, more than two-thirds without workers' compensation — carries forward five years virtually unchanged, while algorithmic systems for extracting labor value grow more refined and harder to detect.

My actual expectation is that none of these three scenarios plays out uniformly across the globe. The most likely outcome is a fragmented future, with bull-adjacent protections taking hold in the EU and parts of Latin America, while the US and most of Asia remain in base-to-bear territory. The same Uber driver experiences radically different rights depending on whether they work in Madrid or Houston, Lagos or Mexico City. Algorithms cross borders instantly and operate identically everywhere; labor rights remain stubbornly imprisoned within national boundaries. That specific contradiction — and the ongoing effort to bridge it — will define platform labor governance throughout the late 2020s and into the 2030s.

## Cascading Effects and Recommendations

This issue does not stay neatly inside the category of labor policy. Once we legally accept the premise that an algorithm can perform all the functions of an employer without bearing any employer responsibilities, that logic migrates. It will spread to hiring decisions, credit scoring, insurance underwriting, housing applications, and every other domain where algorithmic systems are replacing human decision-makers. We already live in a world where recruitment algorithms screen résumés before any human sees them, and credit algorithms deny loans without providing meaningful explanations. If we fail to establish legal accountability for algorithmic management in the labor context, we lose the opportunity to set the precedent that will govern every other domain of algorithmic decision-making that follows.

My recommendation is straightforward, and I'd make it to every stakeholder involved in this negotiation. Before debating wage tables or arguing about the length of convention text, concentrate every available political resource on defending a single sentence: "A company that uses algorithms to perform employer functions is an employer." Make that sentence legally binding. Every other labor protection — minimum wages, social insurance contributions, injury compensation — only becomes enforceable once there is a clearly identified party to enforce it against. Governments should adopt the EU employment presumption principle as the non-negotiable floor, not the aspirational ceiling. Labor organizations should concentrate their advocacy firepower on algorithmic transparency rather than spreading thin across dozens of secondary provisions.

For the rest of us — consumers, voters, everyday users of gig economy apps — the sense of helplessness is at least partially an illusion. Every time you choose a platform with transparent driver policies over one that doesn't explain its practices, every time you ask your elected representative where they stand on convention ratification, you are exercising a form of meaningful pressure. The order you place tonight is, in a small but real sense, a vote for which labor model gets to define the economy of the next decade. The ILO is fighting right now over whether a single word — "binding" — survives intact in an international treaty. That word determines whether 435 million people have legal recourse or just a piece of paper. And I believe this is the most consequential AI governance decision of our current moment, even if most people will never read a sentence about the negotiations taking place in Geneva.