A manager at a Midtown Manhattan office building with glass walls is staring at a dashboard. Her direct reports are listed in a grid on the screen, each with a row of colored indications in the colors green, yellow, and red. The colors represent scores produced by a workforce analytics platform that has been tracking, over the last few months, how much time each employe spends in active windows, how fast they reply to messages, how frequently and for how long they attend meetings, and how their output metrics compare to the team average and to the same time last year. This system was not created by the manager. The scoring technique was not written by her. She utilizes it in the same manner as she does the quarterly sales report: as data that provides her with knowledge about what’s going on, whether or not it is reliable.
Algorithmic management is no longer limited to delivery routes and warehouses. The real-life image of Amazon employes packing boxes while being watched by surveillance devices that precisely time their restroom breaks caused legitimate indignation. Since then, white-collar business settings have adopted the same fundamental logic, which has been translated into software solutions promoted with titles that imply analytical sophistication rather than surveillance.
HR technology platforms are recording digital behavior in a variety of settings, including financial services firms in lower Manhattan, media companies in Hudson Yards, and consulting firms in the Flatiron district. These platforms then translate the data into scores that impact decisions about who gets promoted, who is placed on a performance improvement plan, and who is placed on a list during the quarterly restructuring discussion.
Because the language used by firms to describe these technologies is typically intended to obfuscate rather than to elucidate, it is important to grasp the mechanics. Phrases like “workforce analytics,” “people intelligence,” and “performance optimization” can be found in HR presentations and product marketing materials, but they don’t really explain what the system is measuring or how it turns those measurements into recommendations that can be put into practice.
Email volume, response latency, active hours in tracking software, attendance at calendar events, and keystrokes recorded by monitoring software are examples of proxies for labor that the systems commonly measure. These can be measured. The vendors are interested in providing an affirmative response to the question of whether they have a meaningful correlation with actual work performance, but independent researchers have found this to be much more difficult.
Compared to most jurisdictions, New York City resolved this issue more quickly. Local Law 144, which went into force in 2023, mandates that businesses that use automated employment decision tools for screening or hiring perform yearly bias audits and inform candidates and staff when these tools are being used. It’s a significant regulatory move, the first of its sort at the local level in the US, and it established actual compliance requirements for businesses who had previously used these systems covertly. The picture of enforcement is still evolving. Employes must be informed, but it’s unclear if they have any real options if they think the system gave them an unfair score.
The training issue is the responsibility gap that labor experts consistently highlight. In most firms, managers who receive ratings and recommendations from algorithmic systems are not educated to critically assess those systems. Instead of questioning the software, they are taught how to use it. When a dashboard informs a manager that an employe is in the lowest quartile of engagement metrics, the manager is usually ill-prepared to inquire about the validity of the metric, the suitability of the comparison group, or whether the score represents an aspect of the employee’s work or work style that the software is unable to assess. Even if the data is a proxy measuring something close to what it purports to measure, the advice still has the authority of the data.
Since the phrase “restructuring” is most frequently used in relation to algorithmic-influenced termination occurrences, it merits consideration. When a business declares that it is eliminating 15% of a division as a result of a “data-driven workforce review,” the term effectively conveys impartiality, optimization, and inevitability. Instead than portraying individual job losses as decisions made by individuals using tools that incorporate presumptions about what constitutes good performance, it portrays them as the result of an objective analytical process. Because doing so would require revealing the technique, which the vendors treat as secret and which the corporations claiming it as proprietary have no particular reason to uncover, the validity of the assumptions is rarely investigated openly.
