Artificial intelligence is often presented in binary terms: it’s either the great liberator from human drudgery, or the relentless gravedigger of employment. Yet behind this debate lies a more insidious reality that I call the paradox of induced sophistication. This phenomenon means that any productivity gain is not converted into free time, but immediately reinvested into increased complexity. Far from simplifying our structures, AI is about to saturate the space it frees up, transforming yesterday’s sobriety into a jungle of branching systems and superfluous data.
Parkinson’s Law in the Age of AI
On the employee side, Parkinson’s Law — which states that work expands to fill the time available — applies perfectly. When AI reduces a four-hour task to a few minutes, the human doesn’t choose inaction. On the contrary, they use this efficiency gain to “push the walls” of their deliverable, sometimes straying from their core business.
A simple Excel spreadsheet, once sufficient to manage a project, morphs into a complex application, fed by AI-generated code, covering marginal use cases that would never have been explored before. An app to manage your accounting? Vibe-coded, even though proven and maintained alternatives already exist.
The real efficiency gain becomes zero: you don’t finish earlier, you simply produce a denser, more branching, and often more fragile system.
Management and the Illusion of Omniscience
The top of the pyramid is no exception. Carried by the promise of “infinite intelligence,” management changes the nature of its demands. The Information System is no longer asked to answer a specific question, but to “know everything about everything.”
This ambition, often disconnected from what is reasonable and useful, drives the indiscriminate accumulation of data. Since AI can process petabytes, management demands that every bit of background noise be analyzed. With the risk that by trying to see everything, you end up distinguishing nothing.
The Erosion of Technical Sobriety
From a purely technical standpoint, this transition marks the end of a certain elegance, one imposed by the limits of our human constraints. We are witnessing the replacement of lean methods with a dependency on cumbersome probabilistic systems. This is the era of the “digital Goldberg machine”: building complex machines to accomplish what were once trivial actions.
The cost of “almost”: Where a local spell checker worked with a few megabytes of dictionary and fixed grammar rules, we now deploy massive language models (LLMs) with billions of parameters. We use computing power capable of simulating protein folding just to make sure an email has no typos.
The offshoring of simple intelligence: Searching for a pattern in text used to take one line of code (Regex) executed in microseconds on your processor. Now we send that same text to the other side of the world, to an air-conditioned data center, for the same result.
The use of AI makes simple tasks dependent on systems of ever-growing complexity, turning each micro-task into a heavy transaction.
The Lure of Technological Liberation
History teaches us that technological disruptions rarely produce sobriety or simplicity.
In the 20th century, faster transportation didn’t reduce our commute time; it simply stretched the distances we travel. This is illustrated by Marchetti’s constant: since the Neolithic era, humans have spent an average of one hour per day commuting. If the engine goes faster, we don’t gain rest time — we simply choose to go further.
Similarly, the arrival of household appliances — washing machines, vacuum cleaners, gas stoves — didn’t free humans from domestic chores. It was the standards of cleanliness and social expectations that adjusted upward, instantly absorbing the time freed by the machine.
Later, the introduction of computers in the 1980s paradoxically triggered an explosion in paper demand: the ease of generation led to the multiplication of unnecessary iterations, made trivial by the speed of the printer.
This phenomenon, well known to sociologists as the rebound effect, is a constant: every resource gain (time, energy, space) is immediately reinvested to increase power, precision, or social status. Today, the demand induced by artificial intelligence is about to offset its own productivity gains. Like an extra lane on a highway that, far from easing traffic, ends up attracting more vehicles, AI will saturate our time through an artificial complexification of tasks.
Conclusion: Efficiency Through Temperance
Faced with this mechanics of excess, true intelligence will not lie in technical sophistication, but in our ability to preserve clarity. The challenge is no longer to chase infinite optimization, but to aim for a just measure: doing more than before to meet new challenges, certainly, but never doing more than necessary.
Those who can distinguish ends from means will come out ahead. Using AI for AI’s sake is the shortest path to needlessly complicating your business and making it dependent on LLM outputs. But those who use just the right dose of AI to address their real business needs will gain a strategic edge over the rest.
Listen to this 7m minute reading text transformed into a 22 min podcast using Notebook LLM.
Or watch it as an explainer video
