Automated Decisions, Human Consequences

Ride-hailing and delivery platforms built a particular kind of workplace. There is no office, no shift manager, and often no named person a worker can appeal to. Instead, software assigns jobs, tracks location and timing, scores performance against customer ratings and acceptance rates, and in some cases restricts or ends a worker's access to the platform when those scores fall below a threshold. For years this ran with very little outside scrutiny. That has changed, and the change has implications well beyond the platforms it started with.

What Algorithmic Management Actually Does

Algorithmic management covers four overlapping functions: scheduling and job allocation, continuous monitoring (location, speed, idle time, acceptance and cancellation rates), rating and scoring (often combining customer feedback with system-generated metrics), and consequence, which can range from reduced job offers to suspension to permanent deactivation. The defining feature is not that a computer is involved. It is that decisions with real financial and reputational consequences for an individual can be generated, and sometimes acted on, with minimal or no meaningful human involvement at the point the decision is made.

Workers have repeatedly reported difficulty finding out why a score dropped, why a job stopped being offered, or why an account was suspended, and have struggled to find anyone with the authority to explain or reverse the decision. That opacity, rather than the use of software as such, is what has drawn regulatory and legal attention.

Regulatory and Legal Scrutiny in the UK and EU

In the UK, the Supreme Court's ruling in Uber BV v Aslam established that certain platform drivers were workers rather than genuinely self-employed contractors, which brought them within the scope of protections, including minimum wage and working time rights, that platforms had argued did not apply. That case was about employment status rather than algorithms specifically, but it set the backdrop against which later disputes over automated monitoring and deactivation have played out.

Separately, UK and EU data protection law already gives individuals a right not to be subject to a decision based solely on automated processing where that decision has a legal or similarly significant effect on them, along with a right to obtain human intervention, to express their point of view, and to contest the decision. Worker advocacy groups have used this right as the basis for legal challenges in European courts against ride-hailing platforms, arguing that some deactivation decisions were effectively automated with inadequate human review built into the process. The details of individual rulings vary and continue to develop, but the underlying argument, that a person facing loss of income deserves a real human check before that decision takes effect, has been consistently at the centre of these cases.

The EU has since legislated directly on this. The Platform Work Directive, adopted in 2024 and due to be transposed into national law across member states, requires digital labour platforms to be transparent about the automated systems they use to monitor, evaluate and make decisions about workers, restricts certain fully automated decisions such as account restriction or termination, and gives workers rights to explanation, to human review, and to contest significant decisions. It is the clearest statement yet, in binding law, that the review requirement is not a nicety but a legal obligation.

Why This Reaches Beyond Gig Work

The temptation is to treat this as a gig economy story: unusual employment structures, unusually direct algorithmic control, therefore an unusual problem. We think that reading misses the point. The EU AI Act classifies AI systems used for recruitment, for allocating tasks, and for monitoring or evaluating the performance and conduct of people in employment relationships as high-risk, regardless of industry or employment model. That classification exists because the underlying dynamic is general, not specific to platforms.

Any organisation that uses an algorithmic or AI-assisted system to score performance, flag disciplinary issues, recommend scheduling, screen candidates, or inform decisions about promotion, warning or dismissal is running the same risk that gig platforms ran: a system that produces a confident-looking output, a manager under time pressure who treats that output as the answer rather than as one input, and a worker with no practical route to challenge what happened. The employment structure differs. The exposure, legal and otherwise, does not.

What Makes Human Review Genuine Rather Than Cosmetic

A human signature on an automated recommendation is not the same thing as human review. Review that actually functions as a safeguard tends to share a few features: the reviewer has access to the basis for the system's output in terms they can understand, not just a score; the reviewer has real authority and time to disagree with the system, rather than a target that rewards speed over scrutiny; the person affected can find out that a decision was made, why, and how to contest it; and the organisation keeps a record of how often reviewers actually override the system, since a review step with a zero or near-zero override rate is doing no real work.

None of this requires abandoning automated systems, which can be genuinely useful for surfacing patterns a person would miss. It requires treating the human review step as a real control to be designed, resourced and tested, not as a label attached after the fact to satisfy a policy or a regulator. The gig economy made this visible because the consequences were sharp and the workers affected were numerous and organised enough to take cases forward. The same exposure sits quietly inside any organisation using AI to make or heavily inform decisions about individual employees or workers, whether or not anyone has yet brought a claim.

Sources

• Uber BV and others v Aslam and others [2021] UKSC 5, UK Supreme Court ruling on the employment status of platform drivers.
• Directive (EU) 2024/2831 of the European Parliament and of the Council on improving working conditions in platform work, including provisions on transparency and human review of algorithmic management systems.
• UK GDPR and Data Protection Act 2018, Article 22, on rights relating to solely automated decision-making with legal or similarly significant effects, including the right to human intervention.
• Regulation (EU) 2024/1689 (the EU AI Act), Annex III, classifying AI systems used for recruitment, task allocation, and monitoring or evaluation of workers as high-risk.
• Legal proceedings brought in Dutch courts by worker advocacy groups, including Worker Info Exchange and the App Drivers and Couriers Union, against ride-hailing platforms concerning automated account deactivation and GDPR Article 22 rights.