Your Boss Is Now AI: How Algorithms Manage Millions
Summary of the video “Il tuo capo è l’AI: Come l'algoritmo è diventato il manager di milioni di persone” by Geopop.
Algorithmic management—automated systems that assign tasks, monitor activity, evaluate performance, and issue sanctions—already affects 76% of Italian companies and three out of four workers across sectors from logistics to healthcare to journalism. While these systems can reduce human bias, they operate opaquely, create chronic anxiety, redistribute power to workers' detriment, and lack meaningful human oversight or appeal mechanisms. European regulation is emerging but lags far behind deployment.
The Scale of Algorithmic Management Today
Adoption rates across major economies
Algorithmic management tools—any software automating traditional manager functions like task assignment, monitoring, and performance evaluation—are already embedded in the majority of companies. Italy leads adoption at 76%, Europe averages 79%, and the US reaches 90%, according to a 2025 Oxe survey of over 6,000 managers across six countries.
Types of tools deployed in Italian companies
In Italy, task assignment systems are most common (64% of companies), followed by monitoring systems, with performance evaluation tools least widespread at 27%. These represent different layers of algorithmic control over the worker's day.
Not a future phenomenon—it's present now
Algorithmic management is not an emerging trend but an already-embedded reality. For riders and delivery drivers, the system is explicit and immediate; for office and traditional workers, it is often mediated through platforms that appear like normal work tools but operate on identical underlying logic.
How Algorithmic Management Works: Five Levers
The five-lever structure
Algorithmic management systems across all sectors share a recognizable architecture: assignment (system decides tasks and pace), monitoring (constant data collection on every action), evaluation (data produces scores determining opportunities), gamification (rewards and rankings using slot-machine-like psychology), and automatic sanctions (warnings, suspensions, terminations without human intermediary).
Sector Deep Dives: Where Algorithms Manage
E-commerce and logistics warehouses
The most documented sector. Systems track every operation—picking speed, break times, compliance, production quotas. Workers receive automatic notifications for not meeting pace; accumulation of notifications can trigger automated firing without human manager review. Internal documents from major players reveal systems generating warnings and terminations without supervisor input.
Ride-hailing and delivery platforms
Algorithms place orders, determine rates in real time, build worker scores, and decide access to best opportunities. Account deactivation (equivalent to dismissal) happens automatically without warning, explanation, or human contact. A New York survey found 76% of ride-hailing drivers whose systems were disabled had never received a warning despite high ratings and no violations.
Call centers
Among the first traditional work environments to adopt advanced algorithmic management. Systems transcribe conversations in real time, detect tone changes, and display prompts to guide responses. If customer sentiment drops below a threshold, supervisors receive automatic alerts. The critical flaw: algorithms evaluate worker empathy without access to call context, customer history, or conversation difficulty.
Recruitment and candidate screening
Most large companies use automated resume screening producing compatibility scores. Some use automated video interviews analyzing tone and facial expressions. A major tech company was forced to abandon its AI screening system in 2018 after discovering it systematically penalized resumes with female-context references (e.g., 'captain of women's team') because it was trained on historically male-dominated candidate pools. The algorithm amplified human bias invisibly.
Healthcare
Algorithmic management takes three forms: shift scheduling, patient waiting-list prioritization, and alert systems for clinical staff. Benefits are real—continuous remote monitoring enables timely interventions and reduces unnecessary hospitalizations. However, a 2019 Science study found a widely used algorithm in American hospitals systematically underestimated Black patients' severity compared to white patients because it was trained on historical health spending data reflecting prior disparities in care access. Doctors and nurses must decide whether to trust algorithmic recommendations quickly, often without understanding how they were produced.
Journalism
Algorithmic management is least visible but most pervasive. Algorithms write articles, decide which articles are read (indirectly determining coverage), and shape editorial decisions through real-time analytics showing read counts, duration, and traffic sources. Social media distribution algorithms decide which articles reach audiences. Editorial line is increasingly decided by algorithms, not editors or journalists.
The Human Cost: Paradoxes and Psychological Effects
The autonomy paradox: flexitation
Platform workers formally enjoy freedom—choosing when to work, where to go, hours to dedicate—but this freedom is structurally constrained. Workers who don't work enough are penalized in order distribution; those who refuse too often lose access to best time slots. Researchers call this 'flexitation': exploitation under the guise of flexibility. Freedom is real; the consequences of using it are punitive.
Chronic structural anxiety
Research documents a condition of chronic structural anxiety among workers in algorithmic systems: every action is monitored, but rules are never fully transparent. Workers don't know exactly what rewards or penalizes them. The object of anxiety is not a human interlocutor but an unclear, opaque system. This psychological effect is increasingly documented in both platform and traditional work.
Managers don't understand their own tools
A surprising finding: managers themselves—not just workers—express concern about algorithmic systems. They struggle with unclear responsibilities, difficulty following tool logic, and inadequate worker health protection. Many managers must make decisions based on algorithmic suggestions without fully understanding the underlying logic.
Documented Failures and Biases
The Amazon recruiting scandal
A major tech company abandoned its AI screening system in 2018 after discovering it systematically penalized resumes containing references to female contexts (e.g., captain of women's team). The algorithm had been trained on years of resumes from overwhelmingly male candidates, making historical human bias invisible and amplified.
Healthcare algorithm bias in patient severity assessment
A 2019 Science study analyzed a widely used algorithm in American hospitals for identifying high-risk patients. It systematically underestimated the severity of Black patients' needs compared to white patients because it was trained on historical health spending data that reflected previous disparities in access to care.
The Dutch welfare fraud scandal
The Netherlands government used an algorithm to detect fraud in the child benefit system, leading to unjust fraud accusations against nearly 35,000 families, mostly migrants, over 6 years with severe economic consequences. The scandal caused a political crisis and the resignation of the entire government in 2021. A subsequent 2025 experiment to build a fair fraud-detection algorithm was abandoned after biases persisted despite corrections.
Professions at Risk: The Near Future
Teachers and trainers
Adaptive learning platforms already use algorithms to personalize student learning journeys. The next step, already tested in some US systems, is to evaluate teachers based on algorithmically measured student results. This is not yet widespread in Italy.
Social workers and welfare workers
Several European countries already use algorithms to prioritize cases. The documented risks are severe, as shown by the Dutch welfare fraud case.
Journalists
A 2025 Microsoft report identifies journalists among professions with the highest degree of overlap with current AI capabilities in terms of tasks performed.
Paralegals and legal professionals
Legal analytics systems already analyze case law, produce draft contracts, and identify patterns. The next step is algorithmic evaluation of professional efficiency.
Financial analysts
Algorithmic trading systems have existed for decades. What is changing is that algorithms now evaluate the people operating those systems in real time on standardized performance metrics.
Doctors and nurses
Not for clinical diagnosis, but for time management, patient prioritization, and evaluating prescribing efficiency compared to algorithmic guidelines.
Regulatory Landscape and Protections
What exists: European directive on platform work
A European directive on platform work comes into force in December 2024 and introduces the presumption of employment for those working under significant algorithmic management. Member States have until December 2026 to implement it.
Italian transparency decree (2022)
Italy's 2022 transparency decree requires companies to inform workers about automated systems influencing decisions affecting them. This is an obligation of information transparency, not yet a right to challenge algorithmic decisions.
Italian union protections
Italian unions have negotiated an agreement requiring human representation at all major disciplinary decisions. This is partial protection but cannot be taken for granted.
Critical gaps in protection
No enforceable right exists to receive an understandable explanation of significant algorithmic decisions affecting workers. There is no general obligation for human supervision before automatic sanctions. There is no mandatory independent auditing system for algorithms in high-impact work environments. While 63% of companies report consulting workers about algorithmic management introduction, these consultations typically involve managers, not the workers actually managed by those systems.
The Core Question
Transparency and accountability as the dividing line
The difference is not between algorithms and humans—human managers are influenced by bias, prejudice, and arbitrariness; algorithms can measure fairly and consistently. The real difference is between power exercised with transparency and responsibility versus power exercised opaquely. A transparent, contestable algorithmic system subject to independent auditing with human oversight could function better than an untrained, unaccountable human manager. But an opaque, unchallengeable system making irreversible decisions with no real appeal adds new risks to existing ones.
The unanswered question
Algorithmic management is not the future of work—it is the present, already affecting three out of four Italian companies. The question that must continue to be asked is: who has the right to understand how it works and who has the power to change the rules when it doesn't work fairly? For now, no one has that answer, not even the algorithm.
Notable quotes
The difference isn't between algorithms and humans, it's between power exercised with transparency and responsibility and power exercised opaquely. — Mia Ceran
Algorithmic management isn't the future of work, it's the present and already affects three out of four companies in Italy. — Mia Ceran
Who has the right to understand how it works and who has the power to change the rules when it doesn't work fairly? — Mia Ceran