When Does Efficiency Begin to Hide the Work Required to Produce It?
Efficiency is never only a number
Efficiency is often presented as a simple relationship between input and output. A system produces more with fewer resources, less time, or less visible effort.
But every efficiency calculation begins by deciding what counts as an input.
If a company measures only completed tasks, it may ignore training, correction, maintenance, emotional support, and coordination. If an AI system measures only generated answers, it may ignore data preparation, human evaluation, computing infrastructure, energy, and the people who correct its failures.
Efficiency is not necessarily false. It may be incomplete because the work required to produce it has been excluded from the account.
Marx and the appearance of self-producing value
Karl Marx’s analysis helps explain why productive systems can make their own social conditions difficult to see. A machine appears to produce more output because it is technically powerful, while the labor, knowledge, and social organization behind the machine recede into the background.
The result can look as though value comes from the object itself. The workplace sees a faster machine, a larger factory, or a more capable system. It may forget that the machine depends on workers who designed, operated, maintained, repaired, and organized it.
This is closely related to Marx’s broader concern with how social relationships can appear as relationships between things. A product or system seems to possess power on its own, while the human relations that created that power become harder to recognize.
Thompson and the time hidden inside efficiency
E. P. Thompson’s account of industrial time shows that efficiency is also a way of organizing human time.
When work moves from task-based rhythms to clock-based schedules, time becomes measurable and comparable. A manager can ask how long a task should take, how many tasks fit into an hour, and why the previous output level cannot be repeated indefinitely.
The time saved by a new method does not automatically return to the worker. It can be converted into a tighter schedule or a higher expectation.
What looks like technical efficiency may therefore contain a social reorganization of time. The workplace becomes faster because workers have less control over the pace and use of their own working hours.
Braverman and the disappearance of practical knowledge
Harry Braverman’s analysis of management focused on the separation between conception and execution. Workers may know how a job is actually done, but management can attempt to extract that knowledge, formalize it, and place it inside procedures or systems.
This can make production more predictable. It can also make the worker’s contribution less visible.
The system records the official process while the worker carries the practical knowledge of exceptions, workarounds, and failures. Once the process appears to run smoothly, the knowledge that keeps it reliable can disappear from the efficiency calculation.
The organization sees a standardized operation. The worker knows how much judgment is still required to prevent that operation from breaking down.
What efficiency measurements leave out
| Visible measurement | Often excluded from the account | Who absorbs the omission |
|---|---|---|
| Number of completed tasks | Preparation, checking, and correction | Workers and support teams |
| Response time | Emotional effort and constant availability | Front-line workers |
| System uptime | Maintenance, monitoring, and repair | Technical and infrastructure staff |
| AI-generated output | Data labeling, evaluation, and human feedback | Distributed digital workers |
| Lower production cost | Energy, environmental damage, and social infrastructure | Communities and future users |
Ghost work behind automated systems
Mary L. Gray and Siddharth Suri’s work on “ghost work” shows how apparently automated digital services can depend on people performing small, fragmented, and often poorly recognized tasks behind the interface.
A system may classify an image, identify a phrase, or produce a recommendation. Human workers may have prepared the training examples, reviewed uncertain cases, corrected errors, and evaluated whether the output was acceptable.
The final user sees a seamless service. The labor has been divided into small pieces and placed outside the visible product.
Antonio Casilli’s research on digital labor develops a related point: artificial intelligence does not simply replace human activity. It reorganizes and conceals forms of human work that remain necessary for the system to function.
The more efficiently the interface hides these contributions, the easier it becomes to describe the system as autonomous and the labor as marginal.
Crawford and the material cost of intelligence
Kate Crawford’s analysis of AI challenges the idea that artificial intelligence is weightless or purely informational. AI systems depend on minerals, energy, data centers, supply chains, engineering labor, low-paid support work, and political decisions about where infrastructure is built.
These conditions rarely appear in the moment when a user receives an answer from a model.
The system looks clean because the material and human costs have been moved elsewhere. Efficiency at the interface can coexist with extraction in the wider environment.
A calculation that counts only speed may therefore describe the performance of the screen while ignoring the world required to keep the screen working.
Pasquinelli and collective intelligence
Matteo Pasquinelli’s work helps place artificial intelligence in the history of collective intelligence. Machine intelligence is not created from nothing. It is built from patterns, practices, classifications, language, and problem-solving that have accumulated through social activity.
When these collective capabilities are reorganized inside a technical system, the system may appear to possess intelligence independently. The social source of the capability becomes difficult to identify.
This does not mean that a model is simply identical to the people whose work contributed to it. It means that the system’s apparent intelligence has a history, and that history includes many forms of human contribution that may not receive recognition or control.
Efficiency can hide dependency
The better a system performs, the less visible its supporting labor may become. Successful maintenance prevents disruption. Care prevents complaints. Evaluation prevents obvious errors. Data preparation allows the model to respond. Infrastructure remains invisible while it continues to operate.
This creates a paradox: efficiency can increase dependency while reducing the visibility of the people who sustain that dependency.
When those workers disappear from the account, the system can be treated as cheaper than it really is. The organization may then demand further reductions from the labor that remains visible.
The real question behind an efficient system
To understand efficiency, we must ask what the measurement excludes.
Who prepared the conditions of success? Who absorbs the interruptions? Who performs the corrections? Who pays for the energy, infrastructure, training, and care? Who receives recognition when the system works, and who receives blame when it fails?
Efficiency becomes a problem when the system counts the result but denies the work required to make the result possible.
TravelIAQ Smart Tip: Whenever a process is described as highly efficient, inspect the invisible inputs. Look for maintenance, preparation, supervision, correction, care, energy, infrastructure, and collective knowledge. A system may be efficient at producing an output while shifting the real work and cost somewhere else.