{
  "id": 13507418,
  "title": "La promesa algorítmica de la meritocracia con trampa",
  "url": "https://urgent.news/2026/10/09/la-promesa-algoritmica-de-la-meritocracia-con-trampa",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-09T11:48:21.000Z",
  "source": {
    "name": "Expansion ES",
    "slug": "expansion-es",
    "url": "https://www.expansion.com/expansion-empleo/desarrollo-de-carrera/2026/10/09/6ac7c416e5fdea4a258b45a0.html"
  },
  "original_language": "es",
  "account": "Algorithmic meritocracy promises to reduce the arbitrariness of performance evaluations, but risks invisibilizing valuable tasks that cannot be quantified. Performance assessments have always been subjective, with bosses reviewing an employee's year, mixing results, memories, intuition and affinity, and ultimately deciding who deserves a promotion, raise or warning. Now artificial intelligence claims to replace part of that arbitrariness with data, but the issue is that it may replace a visible subjectivity with a much harder to detect one: the appearance of precision subjectivity. This was discovered by Alan Chang, co-founder and CEO of Fuse Energy, who turned this idea into a management philosophy. Fuse's semi-annual evaluations are heavily based on AI, and Chang aims to eliminate as much human judgment as possible. The company defines itself as \"the definitive meritocracy,\" where results determine authority. Managers are supposed to forget, compare poorly and tend to overvalue the most recent. A machine can remember what happened in February when December arrives, gather dispersed feedback, detect differences between evaluators and uncover contributions that a boss missed. This is what the tech company UST attempts. With over 30,000 employees, it developed AI tools to reconstruct each professional's history and prepare evaluation drafts from achievements, training, skills or comments. The company says these tools have reduced the time spent reviewing profiles and help combat the \"recency bias,\" the tendency to disproportionately value what happened in the last weeks and forget what happened months ago. Managers still make the final decision, which is crucial when evaluation affects salary, promotion or termination. Maruti Suzuki, a subsidiary of Suzuki Motor and a leader in India's automotive market, followed a similar path. Together with EY, they developed an AI-supported system to evaluate thousands of employees through homogeneous tests, simulations and exercises. The company says the process is faster and more consistent, and once again, the machine complements human judgment, it doesn't replace it. A score of 8.7 seems more objective than saying an employee had a good year. But someone decided what data matters, how much weight each variable has and what exactly good performance means. The AI doesn't eliminate these decisions: it shifts them to who decides which data counts, how they are weighted and what good performance means. Also, we must consider Amazon's case: automated systems in their warehouses use productivity, speed and quality data to control performance and generate warnings when an employee falls below certain targets. It's a precise way to measure a variable. However, processing fewer packages might mean many things: a more difficult zone, a technical problem, helping a colleague or working with more care. In knowledge work, the problem is even greater. Here, contributions that are difficult to convert into a metric appear: the engineer who prevents a failure that never happens, the one who leaves a project in time that was doomed to fail, or the employee who dedicates part of their day to training others. If only what the system can register exists, a company might end up making invisible part of its most valuable work. It's not a marginal phenomenon. A 90% OECD survey of over 6,000 managers found algorithmic management tools in 90% of US and 79% of European companies. 60% of users believe these tools improve their decisions, but nearly two-thirds also point out problems, from difficulty understanding how the system decides to who answers when it's wrong. Europe has decided that this is not just an administrative tool. The AI Regulation considers certain systems used to evaluate workers, decide promotions, assign tasks or influence dismissals as high risk. The question is not whether a machine can judge better than a boss, but which judgments are worth converting into algorithms and which would be dangerous to leave out. Automating memory, asking AI to detect contradictions between evaluators, or warning that a boss is ignoring relevant information may be justified. But deciding how much a person is worth is much more debatable. The \"definitive meritocracy\" probably doesn't consist in removing humans from the equation, but in forcing them to evaluate better: using the machine to make visible what managers don't see, without allowing only what the machine can measure to exist.",
  "summary": "Medir el rendimiento laboral con máquinas reduce la arbitrariedad directiva, aunque amenaza con invisibilizar tareas valiosas imposibles de cuantificar. Leer",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}