Op-Ed: Fully automated imbecility, aka algorithmic management, the new workplace hell

The instant widespread acceptance of algorithmic management is a new abyss in the workplace. It’s an evolution of the bureaucratization of HR, hiring, firing, recruiting, KPIs, and other absurdities, extends management practices started in the 1990s.

This almost totally non-responsive business culture, a minefield of performance parameters and “productivity initiatives,” has clicked its way into actual work.

Pioneered in the notoriously hostile gig economy, algorithmic management extends pettiness to all aspects of work in the name of “productivity”.

The core functions of algorithmic management are comprehensive, taking management roles at supervisor level and in some cases middle management:

“Time off task” tracking. A system tracks activity on specific tasks and notes time off that task.

Computer activity logging: This is an old “keystroke monitor” function, compiling data on the simple acts of performance.  

Just-in-time shift scheduling: Needs-based shift scheduling or cancelling of shifts.

Hiring and termination: AI conducts or participates in video hiring interviews and can also fire employees, based on performance metrics.

Performance metrics: Performance metrics may be any discretionary mix of statistics based on core task options, like number of phone calls, task turnover, and case-specific requirements of a job.

What’s managing who? Who’s managing what?

There’s clearly not much in design terms regarding any other human interaction in these functions. Weaknesses in the algorithmic systems are obvious.

In 2025 the OECD flagged an urgent need to address issues like privacy, discrimination, and critically, lack of transparency. That perspective is actually the event horizon of a much deeper issue.

These systems are inevitably managed by people with varying levels of skills and familiarity with algorithmic management. Privacy, discrimination, and transparency are directly related to labour laws in many countries. Ethics? Maybe, maybe not.

Even in the horrific workplace “employment at will” environments of US workplaces, statutory law has the final say. Civil law also has a major role. It’s asking a lot of a middle manager to both trust algorithmic metrics and do their own jobs.

The opportunities for big, expensive mistakes are almost endless:

The metrics may simply be wrong.

Metrics may misrepresent work values, devaluing high-value work in the name of arbitrary things like KPIs. For example, ten 2-minute phone calls which deliver no business may rate higher than a single half-hour call that delivers a new client worth millions.

This type of statistic can’t show that information. The business information will be on the books somewhere, but not on performance metrics. It’s the level of superficiality that’s the problem.

A job is never “just a job”. With any job comes the collateral of human interactions, extra work not being measured, workplace dysfunctions, and imponderables like multi-party friction in the workplace. How do you measure bullying? Can you schedule it?

How do you measure privacy without a court? Even the theory of AI workplace privacy is full of holes. Consider Amazon’s infamous rest room breaks metrics.  Leaving out the issue of hygiene, what about physical realities? What about pandemics, health and safety, crime, workplace hazards, etc.?

Is there any known instance of algorithmic management being able to manage these realities? This is the cutoff, and it’s potentially lethal.

Management means managing people and real-life situations.

The purely algorithmic system is liable to break down in far too many obvious ways. Anyone who’s ever managed “everything” knows that people are always case-specific. One size can’t and won’t fit anyone.

The more restrictive management practices are, and they can easily be self-restrictive, therefore the less effective management can and will be. A truly bone-lazy manager can rely on algorithms until the algorithms cease to apply to situations. A manager trapped in algorithmic data sets may have to invent a way of reporting the facts. Accountability soon comes home to roost like a vulture in both cases.

There’s another issue. Ever since the introduction of workplace surveillance, the reaction to it has been to see it as hostile. Workplace relations are ephemeral enough without added automated stress.

This is where algorithms become major liabilities. Algorithmic management is no excuse for anything at all. The responsibility of management is conducting business properly, not being a mildly interested spectator.

The AI interpretation of “workplace hell”

“Workplace hell” became a default expression roughly the same time as the bureaucratization of HR. There’s now an actual AI definition of workplace hell. You’ll see many familiar terms in that definition.

Even the most insular, aloof, and ignorant “nothing to do with me and my little friends” type of manager should see the dangers. Toxic environments don’t suddenly become non-toxic without management input.

These issues can’t be addressed in any sense by algorithmic management. Even the gruesome “emotional surveillance” software can only see an existing situation, not do anything about it.

This type of surveillance is roughly the equivalent of the old lie detectors. Facial expressions and tone of voice can be totally misleading. Anyone can put on an act. Some people are pretty good at it. Fraudsters and serial killers are often described as very personable.Imbecility barely begins to describe it.

Algorithms can’t manage anything. Managers should cultivate their distrust before disasters.