AI in Business: Measured Progress, Uncertain Economic Balance

IT leaders report results with AI, but few consider them decisive. Between boards of directors that are sometimes more ambitious and a shift towards open-weight models, the economic equation remains uncertain. A BCG survey also highlights incentive biases on the CEO side, while Azeem Azhar describes a "finely balanced" situation.

Signals of Caution and a "Finely Balanced" Situation

According to the Boston Consulting Group, about 70% of CEOs worldwide believe that success in AI is important for the perception of their role. Azeem Azhar mentions a "bear version" of the story, where the use of AI could grow but where the bubble could burst if the generated revenues are not sufficient to finance the necessary infrastructure. In this context, even if results are observed, they do not guarantee the viability of investments. During a discussion with IT leaders, only about eight individuals reported achieving results good enough to interrupt their CEO's summer vacation. Azhar claims he does not have a simple answer to the question of the AI bubble and considers the situation to be "finely balanced."

Concrete Progress, but Still Far from Triumphant Enthusiasm

Azeem Azhar indicates that he posed his questions to nearly 160 IT vice presidents at an event in Las Vegas. When he asked the audience about their ability to achieve measurable results through AI, he found that two-thirds of those present remained standing, a figure higher than he expected. However, he believes that progress remains slow. Some leaders indicate that their boards are becoming more ambitious after initial successes. Even in markets considered slower, such as Italy, Azhar has heard CEOs explain that trust is being built and budgets are increasing despite the encountered mistakes.

Models, Hardware, and Profitability: The Weighty Parameters

Many companies are moving away from expensive cutting-edge models in favor of open-weight alternatives. The lifespan of AI chips in production, cited as ranging from four to eight years, has a significant impact on economic calculations. The central question remains whether the revenues generated by AI are sufficient to finance the massive construction of data centers. Another point to consider is the ability of companies to extract enough concrete value from AI to sustain their investments, preferably at higher rates. To date, there is no documentation of return on investment at the macroeconomic level, as it only manifests through isolated examples. Uncertainty remains regarding the current pace's ability to legitimize the invested amounts.