The paradox of a revolution that accelerates nothing
For the past two years, the global economy has been living inside a double narrative. On one side, companies promise a radical transformation of their operations through artificial intelligence, forecasting productivity gains unseen since the invention of the microprocessor. On the other, the data tell a far more muted story: productivity is flat, measurable gains are limited, and organisations appear to be moving more slowly than before. The gap between the promise and the outcome is widening, as if the most ambitious technological revolution of our time were paradoxically producing a slowdown.
Generative AI has undeniably reshaped the daily routines of millions of workers. It speeds up writing, automates repetitive tasks, and improves the quality of certain cognitive activities. Yet these individual gains have not translated into macroeconomic statistics. Economists speak of a “productivity mirage”: everything seems faster, but nothing truly moves. The phenomenon echoes the 1980s, when personal computing transformed office life without immediately appearing in national accounts. The difference today is that the scale of the promises has created almost instant expectations, which markets have priced in long before companies have been able to extract real value.
This paradox fuels a new tension between narrative and reality. Executives praise the enhanced efficiency of their teams, but employees describe cognitive overload, tool proliferation, and a fragmentation of work that slows processes instead of streamlining them. AI, designed to simplify, often complicates. It adds yet another layer to organisations already saturated with procedures, reporting obligations, and heterogeneous systems. The result is an economy that moves forward while stumbling, as if the most powerful technology of the moment had been plugged into an architecture too fragile to support it.
The invisible slowdown: when organisations struggle to absorb technological power
The core issue is not technological but organisational. AI boosts individual productivity, yet it slows down organisations that adopt it too quickly or too superficially. Companies are discovering that integrating AI is not a matter of swapping one tool for another, but of rethinking entire value chains. Workflows must be redesigned, responsibilities redefined, controls strengthened. Every local gain creates friction elsewhere. AI accelerates execution, but it slows coordination.
This invisible slowdown manifests itself in more meetings, additional validation steps, and human oversight driven by the fear of model-generated errors. Employees spend more time checking than producing. Managers must constantly arbitrate between speed and reliability, in an environment where a single AI-amplified mistake can become systemic. The economy is learning that raw power is not enough: it requires a mental, cultural, and operational architecture capable of absorbing it.
Society also feels this tension. AI promises to free up time, yet it creates a diffuse pressure. Individuals must learn to work with systems that evolve faster than they do, interpret outputs whose logic is not always transparent, and maintain constant vigilance toward tools that can be confidently wrong. Work becomes faster but also more unstable, more demanding, more fragmented. AI does not replace humans; it demands more from them.
Financial markets add another layer of complexity. They value AI as a future engine of growth, yet they impose an adoption pace that exceeds organisations’ real capacity to integrate it. Investment is surging, expectations are soaring, but results are slow to materialise. The risk is a cycle of euphoria followed by disillusionment, not because the technology is overvalued, but because organisations are not yet ready to harness it.
The return of the human: toward an economy where technology reveals its limits
The next phase of this revolution will be less technological than cultural. The companies that truly benefit from AI will be those that re-centre humans not as executors but as arbiters. AI excels at speed, synthesis, and repetition. Humans excel at judgment, nuance, and contextual understanding. Real productivity will emerge from this complementarity, not from substitution.
For companies, the challenge is no longer to adopt AI but to digest it. The most effective organisations will be those that simplify their processes, reduce internal friction, and create environments where technology amplifies collective intelligence rather than dispersing it. AI exposes structures that are too rigid, cultures that are too vertical, and decision chains that are too long. It reveals that speed is not a matter of tools but of architecture.
The economy is entering a phase of maturity in which AI ceases to be a symbol of acceleration and becomes a diagnostic tool for organisational fragility. It forces a rethinking of how we work, collaborate, and decide. AI does not slow the economy by nature; it reveals that the economy was not ready to move as fast as the technology.
The investor and the mirage: between excessive enthusiasm and excessive caution
For investors, AI creates an unprecedented dilemma. On one side, technological euphoria suggests that avoiding exposure means missing the next industrial revolution. On the other, instinctive caution warns against a cycle in which promises still outpace tangible results. Between these two poles, investors must navigate a landscape where the perception of progress advances faster than its economic materialisation.
The first risk is the narrative premium, the valuation boost granted to companies that master the rhetoric better than the execution. Markets have always rewarded compelling stories, but AI amplifies this tendency: a single strategic announcement can trigger a re-rating, regardless of a company’s actual ability to integrate the technology. Investors must distinguish between firms using AI as an operational lever and those using it as a rhetorical one. The line is subtle, but it determines long-term value creation.
The second risk is symmetrical: under-exposure driven by excessive scepticism. Economic history shows that major technological waves reward those who gain early exposure, provided it is disciplined. AI is no exception. Refusing any exposure because productivity gains are slow to appear is to ignore that structural transformations unfold over years, not quarters. Investors must avoid confusing slow integration with lack of potential. The mirage lies not in the technology itself, but in the belief that it will produce immediate effects.
Between these extremes, a middle path emerges: analysing AI not as a sector but as a transversal infrastructure. The winners will not be limited to model developers, but will include companies capable of reorganising their processes, simplifying decision chains, and converting algorithmic power into collective efficiency. The discerning investor watches less for announcements than for subtle signals: governance quality, adaptability, alignment between strategic narrative and operational reality.