Amazon fulfillment-centre productivity discipline

What this models

This example runs on the Peer-percentile-class warehouse productivity discipline network: 12 components and 24 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 2 assumed · 15 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

• assumed

  This models the peer-percentile discipline pattern documented in the Amazon fulfilment-centre productivity and time-tracking case file — the class of deployment where an adverse employment action is generated by threshold arithmetic whose threshold is a share of the governed population. It is not a reconstruction of any Amazon system's code, and nothing here is machine learning: the record describes a measurement stream, a comparison and generated paperwork, and no score or prediction exists in it.

• baseline

  The rate and idle-time ledger is drawn as one store carrying both the measure and the bar because that is what the operator's own description says it is. Amazon's counsel told the U.S. Senate committee that speed-related discipline compares each eligible worker's performance in a given week to the performance of other employees doing the same work at the same facility, and that the slowest five percent may be disciplined; as of 2020 the eligible set was the bottom five percent whose actual rate was 50 percent or less of expectation, and the committee recorded that it does not know whether that threshold remains policy. Amazon's public answer to the California citation describes the same design from the other side: individual performance is evaluated over a long period of time, in relation to how the entire site's team is performing.

• baseline

  The peer edge inside the associate class is drawn at the top rung because within the cohort it governs the mechanism is total and weekly by construction rather than occasional: a share-of-the-site rule names an eligible cohort each week whatever the site's absolute pace. The worker-side quantification comes from a weighted opt-in national survey of 1,484 frontline Amazon warehouse workers fielded April to August 2023, in which 58 percent said their pace is ranked and compared with their coworkers' always or most of the time, against 46 percent of warehouse workers industry-wide. That survey was recruited through targeted advertising with fraud screening and weighted to Amazon's own workforce demographics; it is the best available worker-side measurement and it is not a probability sample.

• baseline

  Two model nodes are drawn because the record documents two triggers sharing one termination surface. The speed and quality ladders are the structured performance review processes Amazon named to the committee. The second path generates job-abandonment terminations out of time and leave state, and the committee found those generated where automated time-tracking failed to account for workers on approved medical leave and registered large negative unpaid-time-off balances. They are kept separate on the diagram and in the copy because they are different triggers, and the live class action pleading the attendance path is a complaint filed in November 2025 whose every clause is an unadjudicated allegation.

• baseline

  The worker-kept shift record is drawn as a first-class record store, and it is the element the catalogue has nothing else like: a record held against the system by the people the system judges. Workers told the committee they run their own parallel timekeeping — 'I keep a timer on my watch to keep track of everything', another carrying a notebook — because they can see and contest the system's own ledger only after the consequence arrives. It is drawn small on every axis except one, because one person writes it by hand while working, and its route into the official record is drawn as the thinnest live check on the board.

• baseline

  The tenure and process-path eligibility gate is drawn as a bounded automated screen because Amazon's own counsel quantified it for the committee: speed-related discipline applies to Tier 1 associates who have worked at least five hours in an eligible process path in the given week and at least 160 hours over their tenure, roughly sixteen shifts, which Amazon described as a minority of the workers at fulfilment centres. It is drawn as a real bound on which workers the generator may reach and as no bound at all on whether a measured shortfall was the worker's to answer for, because the record describes the enrolment rule and describes no correctness screen anywhere.

• baseline

  One reconciliation pathway is drawn at zero because the record documents no audit that performs it, not because the shape looked bare. The unknown-idle-time accumulator renders a conveyor breakdown, a station queue, a pallet problem, a manager conversation and a restroom trip as the same quantity, so the question of why a gap happened is a question the store cannot hold; the only disambiguation the record names is a manager's exemption conversation, which happens on the accrual rather than on the entry. No audit of generated write-ups against cause appears in the congressional record, and Amazon has never published a termination rate, an appeal rate, an override rate or a reversal rate.

• baseline

  Demand reads 3 from documented volume rather than from a default: every task of every eligible associate timed, unknown idle time tracked in some instances down to the second with one published warning citing 97.68 minutes on a named date, about 300 full-time workers terminated for productivity at one Baltimore fulfilment centre across roughly thirteen months in 2017 and 2018 amounting to roughly ten percent of that site's workforce, and 59,017 counted notice violations across two California fulfilment centres between October 2023 and March 2024. The Baltimore figures are site-scoped and eight years old, they are the only per-facility termination base rate anywhere in this record, and Amazon disputed the characterisation of them rather than the count.

• baseline

  Capacity reads 2 because the record documents a working human floor and a narrow one at the same time. It is not 1: a manager held a seek-to-understand conversation and exempted 14 of 48 minutes of accrued idle time, and a Human Resources authority withheld termination from workers who had reached the write-up count, so people with discretion sit on this path and are documented using it. It is not 3: that discretion operates on the exemption of accrued time rather than on the decision, it was bounded in the one instance the committee published, consequences were found arriving more than ten days after the conduct they cite, and no exemption, override, appeal or reversal rate has ever been published by anyone, including Amazon.

• baseline

  No enforcement node and no external-boundary node is drawn, and both absences are structural findings rather than omissions. Nothing downstream of the paperwork appears in this record as a separate action system: crossing a threshold produces the written warning, the final warning or the termination, and the document is the action. No egress is documented either — the measurement, the paperwork and the personnel record stay inside the employer, and what the six state statutes move is a read right toward the worker rather than data toward a third party.

• baseline

  Two reviewer nodes are drawn because the record documents two review functions with genuinely different reach, and each is wired in by its own inbound read. The Human Resources deviation authority is the one internal party that can stop a termination and the same function the record shows as the manual repair channel for leave-state errors before that step was automated; the account of what its power was used for is a manager who would hold back from terminating when headcount was short, with that function's permission. The state labour-standards regulator is the external check that fired twice and whose reach is the written notice of the rule and the worker's access to their own data, never the threshold.

• baseline

  The correlated-reach self-loop is drawn at the middle rung rather than the top, and the reason is a documented scope limit rather than a hedge. One structured process runs per site and per week across a fulfilment network Amazon's own filings describe as roughly 1,300 United States facilities, so a property of the comparison repeats identically wherever it runs. But Amazon told the committee that speed-related discipline applies to a minority of the workers at fulfilment centres — Tier 1 associates in eligible process paths past a tenure gate — so the reach is wide and bounded rather than total.

• baseline

  The sentence at the centre of this record is a quotation from the operator's own counsel that the operator disputes, and both halves travel together everywhere on this board. A letter from an attorney for Amazon to the National Labor Relations Board dated 4 September 2018, obtained and published in April 2019, described a system in which the company's tracking 'automatically generates any warnings or terminations regarding quality or productivity without input from supervisors'. Amazon answered on the record the day the documents were published — 'It is absolutely not true that employees are terminated through an automatic system' — adding that managers can intervene in the process and that terminations can be appealed, and a spokesperson described the same period as about 300 employees of productivity-related turnover at that site. Amazon contested the adjective and not the count.

• baseline

  The congressional findings on this board are the findings of a committee majority that Amazon disputed to the committee, and are carried at that register. The committee's own voice is that 'Amazon uses automated systems to initiate disciplinary procedures' that 'progress in severity and eventually result in termination', and that Amazon's public position that it has no quotas is demonstrably false and semantic rather than substantive. Those are conclusions drawn from internal documents produced to the committee and from two letters by Amazon's outside counsel in 2024. No court and no regulator has adjudicated whether any individual termination on this record issued without a human decision-maker.

• baseline

  This deployment's modelled harm is job loss and an adverse employment action a worker has no documented way to contest before it lands. The injury-and-pace evidence attached to the same buildings — ergonomic-hazard citations, injury-rate analyses, the December 2024 corporate-wide ergonomics settlement — is a different deployment's material and none of it is carried here or computed on this network. The ergonomics settlement appears in this record for one reason and one only: it is the largest governance instrument attached to these facilities and it contains no quota, pace-setting or discipline provision, which is what makes the two deployments separable in the first place.

• baseline

  Three documented channels on this deployment are narrated inside other elements here rather than drawn as separate pathways, and nothing they carried has left the page. The manager’s read of accrued unknown idle time is one act and is drawn once, on the read from the ledger that holds the accumulator, because the record nowhere describes the comparison holding or altering an accrual on its way to a manager. The worker’s own timer or notebook is written and read back by the same person, so it is narrated on the writing edge, and the route that matters for governance — the worker’s account reaching the record that judges them — is drawn in its own right. The daily negative-balance report reaching the Human Resources function is recorded by the committee only in the past tense, as the hand reconciliation an employee worked before that step was automated, so it is narrated on the queue and on that function rather than drawn as a current pathway.

• assumed

  Served people are outside the boundary. The measured associates on this diagram are the operator network — an employer's own workforce and its supervisory chain — and the package recipient appears nowhere. No wage, injury, health or household outcome for any person is computed from anything drawn here; what propagates is institutional error through the workforce, the front-line managers, the Human Resources authority above them and the state regulator outside them. Every figure in the case file is a stipulated or cited property of the deployment, recorded rather than derived.

What this example does not show

• Served people are not on this board. The measured associates here are the operator network — an employer's own workforce and its supervisory chain — and the package recipient is outside the boundary entirely. No wage, injury, health or household outcome for any person is computed from anything on this diagram; what propagates is institutional error through the workforce, the managers, the Human Resources authority and the regulator.
• The sentence at the centre of this case is a quotation the operator disputes, and both halves belong together. A letter from an attorney for Amazon to the National Labor Relations Board dated 4 September 2018 described a system that 'automatically generates any warnings or terminations regarding quality or productivity without input from supervisors'. Amazon answered the day the documents were published — 'It is absolutely not true that employees are terminated through an automatic system' — and said managers can intervene and terminations can be appealed. It disputed the adjective, not the count. No court and no regulator has adjudicated whether any individual termination here issued without a human decision-maker.
• The termination figures are eight years old and site-scoped. About 300 full-time workers terminated for productivity at one Baltimore fulfilment centre between August 2017 and September 2018, roughly ten percent of that site's workforce, is the only per-facility termination base rate anywhere in this record. It is not a network-wide rate, it is not current, and Amazon says the rate of termination is very low without publishing a number.
• The citations are contested and are agency determinations, not judicial findings. California's $5,901,700 citation for 59,017 violations at the Moreno Valley and Redlands fulfilment centres is on appeal; Minnesota's two serious citations totalling $10,500 at Shakopee, one of them the quota-notice count, are contested. No source located for this record reports a resolution of either. The per-facility split of the California total that circulates in secondary coverage is carried nowhere here, because no primary source states it.
• The findings of the congressional committee are the findings of a majority that Amazon disputed to the committee, not adjudicated facts — including that Amazon 'uses automated systems to initiate disciplinary procedures' and that its no-quotas position is demonstrably false. The live class action on the adjacent attendance path, filed in November 2025, is a pleading: every clause about drained unpaid-time-off balances and firing by email is an unadjudicated allegation, and Amazon says claims that it does not follow federal and state law are simply not true. The only merits ruling to date on the quota machinery is a January 2023 dismissal on pleading specificity, which decided nothing about whether the quotas exist.
• The worker-side percentages come from a weighted opt-in survey, not a probability sample. The national survey of 1,484 frontline Amazon warehouse workers was fielded April to August 2023 through targeted advertising with fraud screening and weighted to the employer's own workforce demographics. It is the best available worker-side quantification of the monitoring and the peer ranking, and it is labelled as an opt-in sample wherever a figure from it is used.
• The injury gradient is a different deployment's material and none of it is here. Ergonomic-hazard citations, injury-rate analyses and the December 2024 corporate-wide ergonomics settlement belong to the logistics case that shares these buildings. That settlement appears in this record for one purpose: it is the largest governance instrument attached to these facilities and it contains no quota, pace-setting or discipline provision at all.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

• Amazon's fulfilment centres time every task an associate performs and feed the result into structured discipline processes that generate written warnings, final warnings, and terminations. A letter from an attorney for Amazon to the National Labor Relations Board dated 4 September 2018, obtained and published in April 2019, described a system in which 'Amazon's system tracks the rates of each individual associate's productivity, and automatically generates any warnings or terminations regarding quality or productivity without input from supervisors', and stated that hundreds of workers had been terminated at a single facility between August 2017 and September 2018: about 300 full-time workers at the Baltimore fulfilment centre over that roughly thirteen-month window, representing roughly ten percent of that site's workforce. Amazon disputed the characterisation on the record the day the documents were published — 'It is absolutely not true that employees are terminated through an automatic system' — adding that it would not dismiss an employee without ensuring they received coaching, that managers can intervene in the process, and that terminations can be appealed; a spokesperson described the same period as about 300 employees of productivity-related turnover at that site. Amazon contested the adjective and not the count. Six years later a U.S. Senate committee majority, working from internal documents produced to it and from two letters by Amazon's outside counsel in 2024, found in its own voice that 'When workers cannot keep up, Amazon uses automated systems to initiate disciplinary procedures. These disciplinary procedures progress in severity and eventually result in termination.' That is the finding of a committee majority, disputed by Amazon to the committee, and no court or regulator has adjudicated whether any individual termination issued without a human decision-maker. The tracker behind the 2019 documents was reported as ADAPT, the Associate Development and Performance Tracker; Amazon has confirmed that name in no source verified for this record, and the process names it put on the record with Congress are Structured Productivity Performance Review and Structured Quality Performance Review.

  empirical
  • Investigative Lecher, C. (2019, April 25). How Amazon automatically tracks and fires warehouse workers for 'productivity'. The Verge; host blocked from the verifying environment on 2026-08-28 and not read, with the operative sentence and every figure corroborated by the three records below https://www.theverge.com/2019/4/25/18516004/amazon-warehouse-fulfillment-centers-productivity-firing-terminations
  • Reference Business & Human Rights Resource Centre (2019). Amazon letter shows company uses an automated system to monitor and terminate employees https://www.business-humanrights.org/en/latest-news/amazon-letter-shows-company-uses-an-automated-system-to-monitor-terminate-employees/
  • news CBS News (2019). Amazon under fire for software that recommends firing workers https://www.cbsnews.com/news/amazon-under-fire-for-software-that-recommends-firing-workers/
  • Trade press MIT Technology Review (2019). Amazon's system for tracking its warehouse workers can automatically fire them https://www.technologyreview.com/2019/04/26/1021/amazons-system-for-tracking-its-warehouse-workers-can-automatically-fire-them/
  • Government U.S. Senate Committee on Health, Education, Labor, and Pensions (2024, December 15). The Injury-Productivity Trade-off: How Amazon's Obsession with Speed Creates Uniquely Dangerous Warehouses (Majority Staff Report) https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf

• The rule is a percentile of the governed population, and both sides of the record describe it that way. Amazon's counsel told the Senate committee that its speed-related discipline process compares 'each eligible [worker's] performance in a given week to the performance of other employees doing the same work at the same facility', that the slowest five percent may be disciplined, and that as of 2020 the system identified for potential discipline the bottom five percent of performers whose actual rate was 50 percent or less of the expected rate — with the committee recording that it does not know whether that 50 percent threshold remains policy. Eligibility is gated and quantified: speed-related discipline 'applies to only a minority of Amazon [workers] who work at fulfillment centers, specifically Tier 1 (entry level) [workers] who have worked in an eligible process path for at least five hours in a given week and for at least 160 hours over the course of the [worker's] tenure' — roughly sixteen shifts. Speed runs through Structured Productivity Performance Review and quality through Structured Quality Performance Review, driven by counted defects such as scanning an incorrect item, with a third stream running off 'unknown idle time' or time off task; internal charts reviewed by the committee show speed-related write-ups as by far the most common form of discipline, quality second. Amazon's public answer to the California citation describes the same design from the other side: 'The truth is, we don't have fixed quotas. At Amazon, individual performance is evaluated over a long period of time, in relation to how the entire site's team is performing.' Nothing in this arithmetic is machine learning: there is no learned model, no score, and no prediction in it.

  empirical
  • Government U.S. Senate Committee on Health, Education, Labor, and Pensions (2024, December 15). The Injury-Productivity Trade-off: How Amazon's Obsession with Speed Creates Uniquely Dangerous Warehouses (Majority Staff Report) https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf
  • news ABC7 Los Angeles (2024). Amazon fined $5.9 million for allegedly violating California's warehouse quota law at its Moreno Valley and Redlands warehouses https://abc7.com/post/amazon-fined-59-million-allegedly-violating-californias-warehouse/14972284/

• The documented human step sits at the exemption of accrued time rather than at the decision. Internal Amazon documents provided to the committee track unknown idle time 'in some instances, down to the second' — one warning cites 97.68 minutes of unknown idle time on a named date — and that accumulator fills from conveyor breakdowns, manager conversations, pallet problems, and restroom trips alike. A manager holds a 'seek to understand conversation' about the accrual and may exempt part of it: in the instance the committee published, the manager exempted 14 of 48 minutes for the worker's travel to and from a restroom in a one-million-square-foot warehouse and issued a first written warning for the remaining 34 minutes as a violation of Amazon's Standards of Conduct. The committee found instances of more than ten days between an alleged time infraction and the delivery of the disciplinary consequence, and identified that delay as making it difficult for workers to defend themselves and increasing the likelihood that they are wrongly disciplined. Above the manager sits an escalation authority: a low-level manager reported that management had discretion over whether to terminate workers after a given number of write-ups and would not terminate when headcount was low, but only if Human Resources permitted the deviation from protocol. Throughput is coupled to labour demand as well as to performance — Amazon told the committee that during the peak holiday period 'non-automated warnings, reprimands, write-ups, and improvement plans are paused', and an internal August 2020 chart shows write-ups dipping through the October-to-New-Year peak of 2019 and rising sharply in the first week of January once peak ended. When speed-related write-ups were paused at the start of the pandemic, an internal Amazon team observed that warehouse managers increased their use of other write-up types: behavioural, attendance, and safety. Amazon has published no termination rate, no appeal rate, no override rate, no exemption rate, and no false-positive rate, and says only that 'the rate of termination is very low.'

  empirical
  • Government U.S. Senate Committee on Health, Education, Labor, and Pensions (2024, December 15). The Injury-Productivity Trade-off: How Amazon's Obsession with Speed Creates Uniquely Dangerous Warehouses (Majority Staff Report) https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf

• A second automated termination path runs off a different metric and is where the record shows a human reconciliation step being removed. The Senate committee found terminations for 'job abandonment' generated when automated time-tracking failed to account for workers on approved medical leave and registered large negative unpaid-time-off balances. A Human Resources employee described having, before automation, to review a daily report of workers with negative leave balances — sometimes finding workers on leave with hundreds of thousands of hours of negative time — and personally removing them so they would not be flagged for discipline. One worker recovering from a foot injury was terminated by email a week before her scheduled return. The two paths share the timekeeping and the termination surface and do not share a trigger. The live vehicle on this path is a pleading and nothing in it is adjudicated: Lyster v. Amazon.com Services LLC, a class action over the operator's workplace absence practices filed 12 November 2025 in federal court in New York. Amazon says claims that it does not follow federal and state law 'are simply not true' and that its accommodations team reviews each request individually.

  empirical
  • Government U.S. Senate Committee on Health, Education, Labor, and Pensions (2024, December 15). The Injury-Productivity Trade-off: How Amazon's Obsession with Speed Creates Uniquely Dangerous Warehouses (Majority Staff Report) https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf
  • news CBS News (2025). Amazon sued in class action lawsuit over workplace absence practices (Lyster v. Amazon.com Services LLC, S.D.N.Y., filed 12 November 2025) https://www.cbsnews.com/news/amazon-class-action-lawsuit-warehouse-workplace-absence-practices/

• Whether the governed worker can read the ledger that judges them is contested on the record, and the gap between the two accounts is the case. Amazon's position, stated in response to the California citation, is that 'Employees can - and are encouraged to - review their performance whenever they wish. They can always talk to a manager if they're having trouble finding the information.' The congressional record describes the same channel differently: disciplinary consequences arriving more than ten days after the conduct they cite, and workers running their own parallel timekeeping in response — 'I keep a timer on my watch to keep track of everything', another worker carrying a notebook — because they cannot see or contest the system's own ledger in time. Six state legislatures have since built a statutory version of the same channel, and what they built is a read right rather than a limit on the rule: a written description of each quota and of the discipline attached to it, provided on hire and in the worker's primary language with notice of changes within two business days, plus a right to request the quota description and the most recent 90 days of the worker's own personal work-speed data, answered in New York within 14 calendar days at no cost together with aggregate data for similar workers at the same site. The worker-kept timer or notebook is therefore a record store that exists because the primary one is not legible to its subject in time to contest an entry.

  empirical
  • Government U.S. Senate Committee on Health, Education, Labor, and Pensions (2024, December 15). The Injury-Productivity Trade-off: How Amazon's Obsession with Speed Creates Uniquely Dangerous Warehouses (Majority Staff Report) https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf
  • news ABC7 Los Angeles (2024). Amazon fined $5.9 million for allegedly violating California's warehouse quota law at its Moreno Valley and Redlands warehouses https://abc7.com/post/amazon-fined-59-million-allegedly-violating-californias-warehouse/14972284/
  • Government Assembly Bill 701 (2021), Warehouse distribution centers; California Labor Code sections 2100-2112 https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202120220AB701
  • Government New York State Department of Labor (2023). Quota Requirements, Prohibitions, and Employee Requests Under the Warehouse Worker Protection Act https://dol.ny.gov/quota-requirements-prohibitions-and-employee-requests-under-warehouse-worker-protection-act

• Two independent state labour regulators have cited this design, and both cited the notice rather than the threshold. The California Labor Commissioner's Office cited Amazon.com Services LLC $5,901,700 for 59,017 violations of the state Warehouse Quotas law at the Moreno Valley and Redlands fulfilment centres, the violations occurring between 20 October 2023 and 9 March 2024 and penalised at $100 per violation under Labor Code section 2699(f); the inspection opened on 22 September 2022 and the Warehouse Worker Resource Center assisted. Labor Commissioner Lilia Garcia-Brower stated that 'The peer-to-peer system that Amazon was using in these two warehouses is exactly the kind of system that the Warehouse Quotas law was put in place to prevent. Undisclosed quotas expose workers to increased pressure to work faster and can lead to higher injury rates and other violations by forcing workers to skip breaks.' Amazon appealed: 'We disagree with the allegations made in the citations and have appealed.' Minnesota reached the same conclusion under its own statute: after an October 2023 inspection at the Shakopee facility, Minnesota OSHA issued two serious citations in April 2024 totalling $10,500, one of them for the violation that 'warehouse employees who were expected to meet a quota of selecting, stowing and packaging products were not provided a written copy of the quota before they were expected to meet the quota.' Amazon contested the citations. Both are agency determinations rather than judicial findings and neither was resolved in any source located for this record; the per-facility split of the California total that circulates in secondary coverage is stated by no primary source and is not carried here.

  empirical
  • Government California Department of Industrial Relations (2024). Labor Commissioner's Office Cites Amazon $5.9 Million for Warehouse Quotas Violations (news release 2024-46) https://www.dir.ca.gov/DIRNews/2024/2024-46.html
  • Government Minnesota Department of Labor and Industry (2024). Minnesota OSHA issues citations to Amazon for warehouse distribution worker safety and ergonomic hazards https://dli.mn.gov/node/7026
  • news ABC7 Los Angeles (2024). Amazon fined $5.9 million for allegedly violating California's warehouse quota law at its Moreno Valley and Redlands warehouses https://abc7.com/post/amazon-fined-59-million-allegedly-violating-californias-warehouse/14972284/

• The correction channel the law actually built for this decision surface is a disclosure-and-data right, not a limit on the rule. California Labor Code sections 2100 to 2112, effective 1 January 2022, define a quota as a work standard under which an employee is required to perform at a specified productivity speed or handle a quantified amount of material within a defined time period and under which the employee may suffer an adverse employment action for failing to meet it; require a written description of each quota on hire; bar adverse employment action for failing to meet an undisclosed quota; give employees the right to request the written quota description and a copy of the most recent 90 days of their own personal work-speed data; and create a rebuttable presumption of retaliation for adverse action within 90 days of such a request or complaint. New York's Warehouse Worker Protection Act adds a fourteen-calendar-day no-cost response deadline, notice of quota changes within two business days, provision in the worker's primary language, and a right to aggregate speed data for similar workers at the same site. Six states now carry such statutes — California (2021, effective January 2022), New York (19 June 2023), Minnesota (1 July 2023), Washington (31 May 2024), Oregon (1 January 2025) and Connecticut (signed March 2026, effective 1 July 2026 with notices to current employees due by 1 August 2026). The Senate committee majority's own prescription aims one step further in, and is a proposal rather than law: the No Robots Bosses Act, which would 'prevent employers from exclusively relying on automated systems to make decisions about disciplining or firing workers' and 'require employers using automated decision-making systems to tell workers how the system works and how workers can appeal system decisions', recommended on the committee's finding that 'Amazon subjects workers to discipline based on automated systems that are prone to errors, including firing workers who are on medical leave'; and the Stop Spying Bosses Act, recommended on the finding that Amazon 'closely tracks workers' movements and actions throughout the workday, and uses this information to make disciplinary decisions.'

  empirical
  • Government Assembly Bill 701 (2021), Warehouse distribution centers; California Labor Code sections 2100-2112 https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202120220AB701
  • Government New York State Department of Labor (2023). Quota Requirements, Prohibitions, and Employee Requests Under the Warehouse Worker Protection Act https://dol.ny.gov/quota-requirements-prohibitions-and-employee-requests-under-warehouse-worker-protection-act
  • Trade press Littler Mendelson (2026). Warehouse Quota Notice Laws: Connecticut Joins the Club https://www.littler.com/news-analysis/asap/warehouse-quota-notice-laws-connecticut-joins-club
  • Government U.S. Senate Committee on Health, Education, Labor, and Pensions (2024, December 15). The Injury-Productivity Trade-off: How Amazon's Obsession with Speed Creates Uniquely Dangerous Warehouses (Majority Staff Report) https://www.help.senate.gov/imo/media/doc/amazon_investigation.pdf

• The litigation record on this mechanism is thinner than the volume of coverage suggests, and every item in it stops short of the question. The only merits ruling to date on the quota machinery went on pleading specificity: in January 2023 a U.S. magistrate judge in the Northern District of California dismissed a proposed class action alleging that hourly quotas of roughly 150 to 300 items discriminate against older workers, holding the allegations too vague and writing that 'simply because physical strength declines with age does not automatically mean that older workers are more likely to get injured or fail to keep up with the quotas.' That is reasoning about a discrimination theory and about the sufficiency of a pleading; it is not a holding that the quotas do not exist. The live federal vehicle, filed 12 November 2025 in the Southern District of New York, pleads the adjacent automated absence-control path and is wholly unadjudicated.

  empirical
  • news Wiessner, D. (2023). Amazon beats claim that warehouse quotas are biased against older workers. Reuters, read via syndication because the original was not fetchable https://finance.yahoo.com/news/amazon-beats-claim-warehouse-quotas-193201100.html
  • news CBS News (2025). Amazon sued in class action lawsuit over workplace absence practices (Lyster v. Amazon.com Services LLC, S.D.N.Y., filed 12 November 2025) https://www.cbsnews.com/news/amazon-class-action-lawsuit-warehouse-workplace-absence-practices/

• The governed side of this deployment has been quantified once, in a weighted opt-in national survey rather than a probability sample, and it is labelled as such wherever it is used. The Center for Urban Economic Development at the University of Illinois Chicago surveyed 1,484 frontline Amazon warehouse workers between April and August 2023, drawing respondents from 42 states and 451 facilities, recruited through targeted advertising with CAPTCHA screening, a fake-facility-code trap, and fraud-cluster removal, and weighted to Amazon's 2021 workforce demographics. In it, 77 percent said the technology can tell whether they are actively engaged in their work always or most of the time, against 47 percent of warehouse workers industry-wide; 72 percent said how fast they work is measured in detail by company technology, against 58 percent; and 58 percent said their pace is always or most of the time ranked and compared with the pace of their coworkers, against 46 percent. Asked what the electronic monitoring is used for, 45 percent said it is mainly used to control or discipline workers and 36 percent said it is mainly used to develop workers' skills and abilities, with the control-or-discipline share rising to 52 percent among workers of more than three years. 45 percent said keeping up with the pace of work is hard (47 percent at fulfilment centres against 31 percent at sortation centres), 41 percent feel pressure to work faster always or most of the time, and 53 percent report always or most of the time feeling watched or monitored.

  empirical
  • Academic Center for Urban Economic Development, University of Illinois Chicago (2023). Pain Points: Data on Work Intensity, Monitoring, and Health at Amazon Warehouses (National Survey of Amazon Warehouse Workers) https://cued.uic.edu/wp-content/uploads/sites/219/2023/10/Pain-Points_Final_Oct2023.pdf

• The largest governance instrument attached to these facilities does not reach this decision surface, and that is what separates this case from the pace-and-injury case that shares the same buildings. On 19 December 2024 the U.S. Department of Labor announced a corporate-wide settlement with Amazon requiring Site Ergonomics Leads who review corporate risk assessments and prepare annually updated site-level assessments, together with multiple employee channels — including anonymous ones — for raising ergonomic concerns, across fulfilment centres, sortation centres, and delivery stations in federal OSHA jurisdiction, for a $145,000 penalty. The settlement contains no quota provision, no pace-setting provision, and no discipline provision. The instruments that do reach this decision surface are the state warehouse-quota-notice statutes and the two contested citations issued under them, whose remedy is written disclosure of the rule and access to the worker's own speed data. Nothing in the record adjudicates whether the threshold itself is correct.

  empirical
  • Government U.S. Department of Labor, Occupational Safety and Health Administration (2024, December 19). US Department of Labor announces settlement with Amazon requiring corporate-wide ergonomic measures at facilities nationwide https://www.osha.gov/news/newsreleases/osha-national-news-release/20241219

Where this connects

Institutional pressures in this domain

• Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
• Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
• Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.
• Data & policy drift — The world, the intake process, and the rules change under a system trained on how things used to be — two mechanisms with different remedies: the statistical properties of what the system processes move (concept drift), or the mixture of inputs arriving in deployment differs from the mixture it was trained on (covariate shift).
• Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.

All of them in context on the Hiring & employment screening AI domain page.

Levers available here and the patterns behind them

• Upgrade model — Improve the model
• Gate record entries — Human-in-the-loop write gating
• Mark AI-written records — Provenance labeling
• Check copied records — Reconcile copied records
• Understand the system — Understand the system
• Review on schedule — Oversight cadence & retrospectives
• Assign a challenger — Structured dissent
• Pause AI on alarms — Deployment circuit-breaker
• Store less data — Data minimization

Documented case histories

• The rule with no number to disclose
• A resume screener that learned the past's bias
• Vendor screening across thousands of employers (litigation live)
• Graduate-hiring AI with its audits on the record
• HireVue video assessment (vendor layer)
• The 1959 statute and the integrity video screen (Baker v. CVS Health)
• An internal promotion, a recorded screen, and a captioning request (D.K. charges against Intuit and HireVue)
• Aon pre-hire assessment suite (vendor's own tables)
• The cooperative audit: a paid source-code examination, and what happened to its verdict
• McHire and the 64-million-record custody exposure
• SiriusXM's iCIMS applicant screening
• Checkr gig-economy background screening
• The account goes dark at nine; the reason arrives on day twenty-six
• iTutorGroup Tutor Application Screen
• Meta Job-Ad Delivery: the guardrail and the layer below