Product1 distinct publisher2 min readUpdated
Pandemic-era tracking has been rebuilt into AI-assembled worker records used to predict and discipline. The employer that owns the tooling also owns the timeline someone else will read back to it.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
Nothing in a task-completion feed has to be pointed at one person for it to end up pointed at one person. Duong's managers told her to spend less time with patients after tracking the number and length of her appointments [7]. She said everything was counted, down to the minutes on the phone, and called the standard close to impossible [9]. She went on medical leave, began organising a union with other pharmacists, and says her employment was terminated after the leave [8]. Whatever the merits of that dispute, the sequence it turns on was assembled by the employer's own tooling, and the party that bought the tooling is the party that gets asked to explain each entry.
The pace targets are worth arithmetic. Conveyor-belt scanners in one case monitored workers to ensure they inspected hundreds of items per hour [14]. Two hundred an hour leaves 18 seconds an item; six hundred leaves six [1]. Hayley Tsukayama of the Electronic Frontier Foundation says workers are pushing back on this class of tool on the grounds that timing task completion contributes to injuries [11]. A system that publishes six seconds as a threshold has also documented the threshold.
The remote-work framing around this equipment has drifted from the cases. The three the Associated Press account opens with are a warehouse belt, a clinic and a college writing class, none of which happen at home [2], including an administrator reading the papers students uploaded to a college learning system [13]. Tsukayama's advice, that anyone on a workplace-issued machine should assume some monitoring [12], is a boundary drawn around laptops. It does not cover a scanner on a belt or a shared gradebook, and neither does most acceptable-use language written in 2020.
The undisclosed broker sharing is the part of the stack the buyer never specified and cannot inspect from an admin console [5]. An employer telling staff the data stays internal is making a promise about a supplier's integrations rather than its own systems. Wilneida Negrón's point that workers with the least labour-market power become the testing ground for the most intrusive collection [4] reads, from the operator's chair, as a fair prediction of where the first complaint arrives, and of which business unit will be surprised by it.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Pharmacist Lannie Duong said her employer tracked the length of her phone calls and appointments, and in performance evaluations questioned why she took so long with each patient.
Managers told a pharmacist to spend less time with patients after tracking the number and length of her appointments.
Duong said everything was counted, including how many minutes she was on the phone, and that the minutiae was ridiculous and the tasks felt impossible.
An investigation by Vanderbilt University, Northeastern University and the University of California at Berkeley found that some of the most common workplace monitoring programs have shared names, email addresses and other personal worker data with hundreds of outside data brokers and technology companies without clearly disclosing the practice.
Duong went on medical leave and began volunteering with other pharmacists to organise a union; she says her employment was terminated after the medical leave.
College administrators read an adjunct professor's comments to students on personal essays; while teaching a college writing class, Arianna Anaya discovered that an administrator was reading the papers her students uploaded to the school's learning system.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single wire source, named experts, unverifiable anecdotes
All material comes from one publisher carrying one Associated Press service story. Its strongest documentary anchor, the three-university finding on data-broker sharing, is cited without title, date, methodology or the names of the monitoring programs involved. The two worker cases are first-person accounts with unnamed employers and no employer response, and the warehouse example is a single unattributed sentence. Expert attribution is specific and on-the-record, which lifts the floor, but nothing in the cluster is independently checkable.
Broad persistence asserted, deployments illustrated but unquantified
The source asserts continued use by thousands of employers of pandemic-era tracking tooling and illustrates three distinct deployment settings (warehouse pacing scanners, clinic call/appointment timing, university LMS administrator visibility), plus an academic finding of routine third-party data egress from common monitoring programs. That is real breadth of setting, but there are no product names, install counts, market shares or time series, so the magnitude of adoption cannot be sized from this cluster.
Cluster framing outruns a hedged source
The cluster headline and dek assert that dashboards have become dossiers used to predict and discipline and that ownership of the tooling settles ownership of the record. The source supports the direction but is more hedged: 'dossiers' and behavior prediction are one advocacy-side expert's characterization, the same expert says many employers deploy transparently with employee-board review, and neither worker case establishes that an AI-assembled record caused the adverse outcome. The gap is one of certainty and mechanism rather than of substance, and the data-broker finding gives the ownership framing partial support.
Advocacy-side sourcing, no vendor or employer counterparty
Every substantive voice has a stake in the framing: Coworker exists to help workers organize, the EFF advocates for digital privacy, both named workers describe adverse employment outcomes and one was union-organizing, and the compliance manager works in the privacy and AI compliance market that this risk sustains. The monitoring vendors and the employers in each anecdote are absent, so no counterparty incentive is represented. The service-piece format also rewards actionable alarm. This is disclosed advocacy rather than hidden interest, so it is a substantial but visible tilt.
Directionally credible, specifics weakly held
Confidence is moderate-low. The general pattern, persistent post-pandemic monitoring tooling now feeding evaluative records, is consistent across multiple named experts and three independent settings within the source, and the state-law detail is concrete. But the cluster has one publisher, no corroboration, no vendor or employer response, an unverifiable key study citation, and a truncated body; one derived ledger claim is contradicted by the source's own facts. Individual specifics should be treated as unconfirmed.
product
Two years of jobs data invert the AI risk rankings: clerks shrink, managers grow1 distinct publisher
product
From Omaha students to ICE: the shock glove sold as camera-proof force1 distinct publisher
build
Flock cut retention to seven days and published the number that made 30 indefensible1 distinct publisher
product
OpenAI shipped a teen ChatGPT. The over-65 cohort doubled to 23% and got nothing.1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.