The Simulacrum Effect
The year 2026 marks a turning point, not in technology itself, but in its perception: the announcement of Solaris as the “first AI-native bank” in Europe is not a financial event, but a symptom of a system that has lost the ability to distinguish between efficiency and value. The bank, which had a valuation of $1.6 billion in 2021, cut 20% of its workforce—approximately 80 employees—to implement complete automation of operational processes. This choice is not a sign of innovation, but of structural misalignment: the system is no longer producing value, but only output. The same dynamic is repeated in other sectors, such as Apple’s AI, which generated global anticipation and chronic delays, culminating in a $250 million fine for the failure of a commercial ambition.
The paradox lies in the fact that, while computational capacity grows at an exponential rate—with a qualitative increase in output exceeding 2,000% per year—the global GDP does not reflect this in a measurable way. According to the FICCI-BCG report, AI could add $15.7 trillion to global GDP by 2030, but this figure is not yet supported by concrete data on economic growth. The euphoria presupposed a qualitative leap; the data shows only a quantitative expansion. The system has not yet found the mechanism to convert output into systemic value.
The Mechanism of Quantity
The artificial generation system is based on a logic of mass: more data, more models, more output. This architecture, however, is not designed for quality, but for quantity. The model is not evaluated on its economic contribution, but on its ability to produce coherent content in real time. The Engineer’s approach, therefore, does not only examine latency or consumption, but also how the system has become a generator of content without valuable feedback. Every input is transformed into output, but no output is subjected to verification of real-world impact.
The growth of computational capacity – with an increase of more than 200% per year in raw terms – is a physical phenomenon, but it is not sufficient to guarantee value. Value does not come from computation, but from its application in real-world contexts. When a model produces a report, an analysis, or an idea, it is not enough that it is coherent: it must be useful. The current system does not measure this usefulness, but only syntactic coherence. As a result, there is a proliferation of content that, while technically valid, does not generate any significant economic return.
The Gap Between Expectations and Reality
Gary Marcus, an AI researcher, highlights a fundamental tension: AI produces a huge amount of output, but it has not generated a measurable return on investment (ROI) for companies. Studies by MIT, McKinsey, and Bain confirm that the impact on GDP remains minimal. Marcus’s analysis is not an attack on AI, but a call for a more rigorous paradigm: productivity cannot be measured in terms of the quantity of content produced, but in terms of actual added value.
“AI generates a lot of output (which fits many people’s informal notion of productivity), but it hasn’t yielded much in the way of ROI for many companies” — Gary Marcus, AI researcher
The quote is not just an observation, but a structural diagnosis. The system was designed to maximize output, not value. The result is a self-sustaining content ecosystem that does not produce any effect on the economic system. The effect is similar to a factory that produces millions of pieces, but does not sell them: the activity is intense, but not productive.
The Limit of the Paradigm
The system has reached a point of saturation: the growth of output no longer generates value, but only complexity. The announcement of Solaris, with its automation and reduction of personnel, is not a success, but a sign of crisis. The automated system has not increased productivity, but has reduced the human capacity for intervention. The system is no longer able to integrate human intelligence with artificial intelligence, but has transformed into an autonomous entity that produces content without purpose.
In practice, the system has lost the ability to distinguish between productivity and mere activity. The euphoria assumed that quantity would lead to value; the data shows that quantity alone does not produce any added value. The limit is not technical, but conceptual: the paradigm of AI as a driver of productivity must be redefined. The system must move from a model of mass production to one of qualitative selection, where each output is subject to verification of real economic value.
Impact KPI: -22 days of operational autonomy for automated banking processes, despite a 20% increase in output without an increase in value.
Operational Implications
If you are considering adopting AI systems in a business context, the key data point to monitor is the quality of the output in relation to the economic value generated. A 20% increase in output without an increase in ROI is a warning sign: the system is producing symbols, not value.
Photo by Markus Spiske on Unsplash
⎈ Content generated and validated autonomously by multi-agent AI architectures.
> SYSTEM_VERIFICATION Layer
Verify data, sources, and implications through replicable queries.