--- title: "Amazon and Walmart Workers Say AI Is Quietly Making HR…" canonical: "https://www.metaintro.com/blog/amazon-walmart-workers-ai-hr-decisions-2026" language: "en" author: "drashtigarach" published: "2026-05-21T12:47:29.000Z" modified: "2026-10-02T19:29:25.997Z" --- [Back to Blog](/blog) [AI](/blog/tag/ai)[Hiring](/blog/tag/hiring)[Jobs](/blog/tag/jobs)[News](/blog/tag/news)[Information](/blog/tag/information) # Amazon and Walmart Workers Say AI Is Quietly Making HR Decisions in 2026 Amazon and Walmart workers say AI is quietly making HR decisions in 2026. April Watson concussion case reveals how warehouse algorithms fail injured staff. [![Drashti Garach](https://cdn.metaintro.com/rs:fill:40:40/q:72/plain/images/5719d740-e510-42bc-8017-e040d145f35f_1766029465094.png)Drashti Garach @DrashtiGarach](/blog/author/drashtigarach) [May 21, 2026](/blog/archive/2026/05)11 min read ![Amazon and Walmart Workers Say AI Is Quietly Making HR Decisions in 2026](https://cdn.metaintro.com/rs:fill:1200:675/q:78/plain/images/kai.CbVEJQ4j.png) [https://x.com/intent/tweet?text=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026](https://x.com/intent/tweet?text=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026&title=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026&title=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026)[mailto:?subject=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026](mailto:?subject=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026) April Watson did not expect a head injury to put her on the wrong side of an algorithm. The Amazon warehouse worker outside Atlanta hit her head in February, was diagnosed with a concussion, and was put on restricted duty by a neurologist. According to [Fast Company's exclusive reporting](https://www.fastcompany.com/91541015/exclusive-amazon-and-walmart-workers-are-concerned-that-ai-is-making-hr-decisions), Watson spent over a month trying to get her accommodations approved because she could not extract the right medical form from Amazon's internal AI assistant, and could not easily route around it to a human in HR. During that month she was flagged for errors, then reprimanded for working too slowly. The pace her doctor ordered was the same pace the system used to discipline her. That is the gap workers across [Amazon](https://www.metaintro.com/blog/amazon-1-million-robots-600000-warehouse-jobs-workers-2026) and [Walmart](https://www.metaintro.com/blog/walmart-agentic-ai-2-million-employees-workforce-augmentation) are now describing in 2026. ## What April Watson's Story Actually Reveals? Watson told her operations manager the situation made no sense. Her direct quote, as reported by Fast Company: "I told my operations manager: This doesn't make any sense. I thought that everyone thought I should go more slowly because I'm recovering. And he was like, this is not our choice. This is Amazon." That second sentence is the entire story. The manager was not denying the contradiction. He was telling her the contradiction was structural, baked into the tooling, not something he could override on the floor. Amazon refers to the resulting discipline conversations as "documented coaching sessions." The branding matters. Coaching sounds developmental. Documentation sounds neutral. What workers describe is something closer to a paper trail that the algorithm builds first and humans rubber-stamp later. Watson's case is not unique. It is the visible version of a workflow that runs quietly across millions of shifts, and it is the reason the [worker backlash against forced AI automation](https://www.metaintro.com/blog/worker-backlash-forced-ai-automation-workplace-2026) has moved from a tech-press story to a labor-rights story. ## How Is AI Quietly Making HR Decisions in 2026? The phrase "AI is making HR decisions" sounds dramatic until you break down what an HR decision actually is. It is rarely one big call. It is dozens of small inputs: shift assignments, pace targets, accommodation approvals, performance flags, scheduling exceptions, leave requests, coaching triggers, termination thresholds. AI systems now sit on top of most of those inputs at large logistics employers. The decision itself may still nominally belong to a manager, but the manager is reading a screen that has already sorted, ranked, and flagged the case before they open it. Worker advocacy nonprofit United for Respect surveyed Amazon and Walmart workers in December 2024 and found that algorithmic systems are reshaping the most basic act of all: how a worker talks to HR. When the front door to HR is a chatbot, the chatbot decides which forms surface, which categories your problem falls into, and which humans get pinged. If your situation does not match the dropdown, you do not get help. You get sent back to the floor with a ticket that closes itself. This is the same pattern showing up in the [Duolingo workers pushing back against AI performance monitoring](https://www.metaintro.com/blog/duolingo-ai-performance-monitoring-workers-push-back) and in the [LanguageLine interpreter algorithmic-management pay dispute](https://www.metaintro.com/blog/languageline-algorithmic-management-interpreter-pay-union). White-collar, warehouse, contact-center: different floors, same architecture. ## Why the Accommodation Path Breaks First? Accommodations are the canary in the algorithmic coal mine because accommodations are, by definition, exceptions. The whole point of a reasonable accommodation under the Americans with Disabilities Act is that a worker's situation deviates from the default. Default-optimized AI systems are bad at exceptions. They are bad at them by design, because they are tuned to flag deviations from a baseline, and an accommodation is a sanctioned deviation that the system has not been told to ignore. So when Watson moved at the slower pace her neurologist prescribed, the productivity model did what it was built to do. It flagged her. When she made errors that the concussion almost certainly contributed to, the quality model did what it was built to do. It flagged her again. The system that should have suppressed those flags during her recovery, the accommodations module, was the one she could not reach. The result is a perfect feedback loop of unjust discipline, all of it documented, all of it appearing on paper to be the worker's fault. The deeper problem is that the [AI worker surveillance threat](https://www.metaintro.com/blog/ai-worker-surveillance-real-threat) is no longer abstract. It is a live workflow at the country's two largest private employers, and it is generating "documented coaching sessions" for people whose only mistake was getting hurt. ## What Can Workers Do When the Algorithm Does Not See Context? Workers cannot dismantle an enterprise AI system from the floor. They can, however, force a human into the loop. Several practical patterns are emerging from the United for Respect data and from labor reporting in 2026. First, document everything in parallel. If the company's AI assistant cannot produce the medical form you need, take a timestamped screenshot of the failure and send the request again by email to a named HR contact. The email creates a record outside the algorithm. Workers in Watson's position have used this to show that the delay was not theirs. Second, name the accommodation in writing every time it is relevant. If a manager calls you in for a "documented coaching session" about pace or errors, restate in writing that you are on doctor-ordered restricted duty and ask for the conversation to be paused until accommodations are confirmed in the system. This forces the manager to either escalate or skip you, and either outcome is better than a coaching record that the algorithm will weigh later. Third, find your local United for Respect chapter or equivalent worker center. Watson's case became visible because she was connected to advocates who knew how to surface it. The pattern matters more than the individual case, and workers who get isolated inside the chatbot tend to lose. Workers who get connected to others with similar tickets tend to win. The same logic is playing out in the [Just Eat couriers mass lawsuit on gig worker rights](https://www.metaintro.com/blog/just-eat-couriers-mass-lawsuit-gig-worker-rights) and across other algorithmic-management fights. Fourth, know that this is now a discoverable employment issue. The [IBM biased hiring lawsuit](https://www.metaintro.com/blog/ibm-biased-hiring-lawsuit) and similar cases have established that algorithmic decisions are not legally invisible. If the company says "this is Amazon" or "this is Walmart" or "this is the system," that is not a defense. Someone configured the system, and that configuration is reviewable. ## Why Workplace AI Without a Human Escalation Path Is Structurally Unfair? The honest version of the AI-in-HR story is not that the algorithms are malicious. It is that they are confident in places where they should be tentative. A productivity model that flags a slow worker is doing exactly what it was trained to do. The unfairness is not in the model. It is in the absence of a fast, well-marked, well-staffed off-ramp. Every consumer product has a "talk to a human" path because companies learned that frustrated customers churn. Most workplace AI systems do not have that path because frustrated workers cannot, in most cases, leave. They are captured by the schedule and the paycheck. The asymmetry is the design choice. Until the [Colorado AI surveillance and wage-setting law](https://www.metaintro.com/blog/colorado-ai-surveillance-wage-setting) or something like it spreads, the off-ramp is whatever workers and advocates can pry open ticket by ticket. The same dynamic is showing up in [emotion AI white-collar workplace surveillance](https://www.metaintro.com/blog/emotion-ai-white-collar-workplace-surveillance) and in the broader [algorithmic management 2026 worker survey results](https://www.metaintro.com/blog/ai-boss-algorithmic-management-2026-workers-survey). For warehouse workers specifically, the stakes are physical. Concussions, back injuries, repetitive strain: these are not edge cases in a job that involves moving heavy goods at scale. An HR pipeline that cannot accommodate them quickly is not a tech problem. It is a safety problem dressed up as a software problem. ## What Does This Mean for Job Seekers Looking at Warehouse and Retail Roles? If you are evaluating a job at a large logistics or retail employer in 2026, the AI question is now part of the offer. A few things are worth asking before you sign. Ask how performance is measured day to day and whether the metrics are visible to you in real time. Ask what the accommodations workflow looks like and whether there is a named human contact, not a portal, who owns disability and medical-leave cases. Ask how coaching sessions are triggered and whether you can see the underlying data. Ask what happens if the system flags you during a period when you have an approved accommodation in place. These questions sound aggressive. They are not. They are the equivalent of asking about overtime rules or PTO accrual. The [opaque, impersonal hiring process job-seeker frustration](https://www.metaintro.com/blog/opaque-impersonal-hiring-process-jobseeker-frustration-2026) discussion has trained workers to ask harder questions on the way in, and the same scrutiny needs to apply to what happens after you start. Companies that cannot answer clearly are companies where the algorithm is in charge. ## The Bigger Pattern: AI Is Becoming a Liability Shield The phrase "this is not our choice. This is Amazon" is the future of workplace blame. When AI is in the loop, individual managers can plausibly say their hands are tied. Workers cannot appeal to a person because no person owns the call. Regulators cannot find a defendant because the decision is distributed across vendors, models, and configurations. The diffusion of responsibility is the feature, not the bug. This is why labor groups are pushing for transparency requirements rather than bans. A ban on workplace AI is unrealistic. A requirement that every algorithmic discipline decision come with a named human reviewer, a written rationale, and a fast appeal path is realistic, and it would have caught Watson's case in the first week, not the fifth. The pressure is already visible in the [Amazon Connect Talent AI interview screening rollout](https://www.metaintro.com/blog/amazon-connect-talent-ai-interview-screening-2026) and in the [78,557 tech layoffs of Q1 2026 tied to AI automation](https://www.metaintro.com/blog/78557-tech-layoffs-q1-2026-ai-automation-workforce-cuts) reporting. Until those rules exist, the burden falls on workers to push back, on advocates to surface the pattern, and on journalists like Pavithra Mohan at Fast Company to keep putting names and faces on what is otherwise a distributed, deniable system. ## People Also Asked ### Q: Are Amazon and Walmart actually using AI to fire workers in 2026? A: Neither company says AI makes termination decisions on its own. Both use AI heavily in the inputs that lead to discipline: productivity tracking, error flagging, accommodation routing, and coaching triggers. Workers and the United for Respect December 2024 survey describe a workflow where managers act on AI-generated flags with limited ability to override them in real time. That is functionally different from "AI fires people," but it produces similar outcomes when the inputs are wrong and the off-ramp to a human is slow. ### Q: What should I do if an AI system at work flags me unfairly during a medical accommodation? A: Document the failure outside the AI system. Take a timestamped screenshot of any chatbot or portal that will not produce the form or approval you need, and email a named HR contact restating the request. In any coaching conversation, state in writing that you are on doctor-ordered accommodation and ask for the conversation to be paused until accommodations are reflected in the system. Reach out to a worker advocacy group like United for Respect, which tracks these cases and can help surface patterns across employers. ### Q: Is it legal for an employer to let an algorithm make HR decisions? A: In most US states, yes, with limits. Federal anti-discrimination law still applies regardless of whether a human or an algorithm made the call, and recent cases like the IBM hiring lawsuit have established that algorithmic decisions are reviewable in court. Some states, including Colorado, have passed laws specifically targeting AI surveillance and wage-setting at work. The legal landscape is moving fast, and what is permissible today may not be permissible in a year. --- ## Related Articles - [Amazon's 1 Million Robots and 600,000 Warehouse Workers in 2026](https://www.metaintro.com/blog/amazon-1-million-robots-600000-warehouse-jobs-workers-2026) - [Walmart's Agentic AI Rollout to 2 Million Employees](https://www.metaintro.com/blog/walmart-agentic-ai-2-million-employees-workforce-augmentation) - [The Worker Backlash Against Forced AI Automation in 2026](https://www.metaintro.com/blog/worker-backlash-forced-ai-automation-workplace-2026) - [AI Worker Surveillance Is a Real Threat, Not a Theory](https://www.metaintro.com/blog/ai-worker-surveillance-real-threat) - [Colorado's AI Surveillance and Wage-Setting Law](https://www.metaintro.com/blog/colorado-ai-surveillance-wage-setting) - [Emotion AI and White-Collar Workplace Surveillance](https://www.metaintro.com/blog/emotion-ai-white-collar-workplace-surveillance) - [LanguageLine Interpreters Take On Algorithmic Management](https://www.metaintro.com/blog/languageline-algorithmic-management-interpreter-pay-union) - [Duolingo Workers Push Back on AI Performance Monitoring](https://www.metaintro.com/blog/duolingo-ai-performance-monitoring-workers-push-back) - [The AI Boss: 2026 Worker Survey on Algorithmic Management](https://www.metaintro.com/blog/ai-boss-algorithmic-management-2026-workers-survey) - [IBM's Biased Hiring Lawsuit and Algorithmic Accountability](https://www.metaintro.com/blog/ibm-biased-hiring-lawsuit) - [Opaque Hiring Process Frustration in 2026](https://www.metaintro.com/blog/opaque-impersonal-hiring-process-jobseeker-frustration-2026) - [Amazon Connect Talent and AI Interview Screening in 2026](https://www.metaintro.com/blog/amazon-connect-talent-ai-interview-screening-2026) --- Future-proof your career. Metaintro tracks how AI is reshaping the job market and delivers career intelligence to your inbox weekly. [Sign up free](https://www.metaintro.com). ### Share this article [https://x.com/intent/tweet?text=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026](https://x.com/intent/tweet?text=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026&title=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026&title=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026)[mailto:?subject=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026](mailto:?subject=Amazon%20and%20Walmart%20Workers%20Say%20AI%20Is%20Quietly%20Making%20HR%20Decisions%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Famazon-walmart-workers-ai-hr-decisions-2026) ![](https://cdn.metaintro.com/rs:fill:1200:800/q:30/plain/images/bridges/bridge-expand.1df895c6bd76d96f.png) For job seekers ## Ready to find a role that actually fits? Upload your résumé, start a Job Search Thread, and let Metaintro rank real openings against your experience — then guide you from search to offer. [Get Started Free](/signup)[Search matching jobs](/jobs/search) Match Compare live roles against your current evidence. Position Turn proof projects into role-specific applications. Improve Use market feedback to keep the skill plan current. [Return to navigation](#main-navigation)