Over the past two years, generative artificial intelligence has dominated headlines, promising productivity gains, and endless predictions about its impact on the global economy. Tools like ChatGPT, Copilot, and others are presented as the vanguard of a new technological revolution. However, a recent MIT report, titled "The Generative AI Gap: The State of AI in Business in 2025," has raised alarm bells: what if we are living in the midst of an AI bubble?
El MIT analizó 300 implementaciones públicas de IA, entrevistó a 150 líderes empresariales y encuestó a 350 empleados de diferentes compañías. El resultado es un baño de realidad, y he aquí algunos datos:
Despite investments of between $30 and $40 billion, 95% of generative AI projects fail to generate a measurable financial impact.
Only 5% of custom pilots reach production and deliver a significant return.
Most enterprise uses of AI are concentrated in marketing and sales, while the real savings opportunities lie in automating internal processes and reducing outsourcing.
The conclusion is clear: it's not that the technology is deficient, but that business integration is failing. The models are powerful, but they are hampered by fragile processes, rigid systems, and ill-prepared corporate cultures.
MIT has coined an interesting term: the generative AI gap. It refers to the contrast between superficial enthusiasm and operational reality. Many managers announce AI strategies at conferences and on social media, but internally, hardly any tangible results are achieved.
Meanwhile, employees are adopting AI in parallel. More than 90% of workers use tools like ChatGPT without corporate approval. This phenomenon, known as "shadow AI," shows that individual utility is evident, but institutional utility is conspicuously absent.
In other words, there's a disconnection between what the company officially promotes and what teams actually need to work better.
According to the report, the most successful projects aren't the most eye-catching, but the most practical. AI is proving especially useful for automating internal processes, such as document management, top-level customer service, and reducing dependence on external suppliers.
Furthermore, job substitution isn't occurring en masse within permanent workforces, but rather at the margins: outsourced jobs are the first to be replaced. Many companies aren't laying off employees; they simply stop renewing contracts or filling vacancies with human staff.
This contrasts with the alarmist narrative of a "wave of mass layoffs" and suggests that the transformation will be more gradual and less visible, though just as profound.
But if companies don't achieve clear benefits from AI, it's worth asking: what about the companies that develop it?
Here another paradox arises. Giants like OpenAI, Anthropic, and Cohere are still not profitable. Even Microsoft and Google, despite their progress, rely on long-term strategies to monetize AI. The reality is that most of these companies survive thanks to large investment rounds, hoping that revenue will come later.
This inevitably brings to mind other tech bubbles, such as the dot-com bubble of the 2000s, where many startups grew thanks to speculation rather than results.
And the question that arises is: Are we facing a bubble?
If we compare all this data, the idea of an AI bubble doesn't seem so far-fetched. The symptoms are there:
. Excessive expectations: An immediate revolution is promised that has yet to arrive.
. Million-dollar investments with no return: Billions spent on projects that don't produce profits.
. Dependence on venture capital: AI companies are not sustaining themselves with their own revenue.
. Disconnection between employees and management: The shadow AI phenomenon reveals that workers do see utility, but corporate structures fail to capitalize on it.
The key question is not whether there is a bubble, but how long it can be sustained. Are we facing an imminent correction, or will it be a slow adaptation in which only the most useful applications survive?
Like any great innovation, AI will inevitably go through a cycle of enthusiasm and adjustment. Let's not forget that the same thing happened with the internet, smartphones, and the cloud. Early promises almost always outpace initial results, but over time, practical uses become established.
What will be interesting will be to see which AI applications manage to integrate into the daily lives of businesses. These probably won't be marketing chatbots or catchy text generators, but rather invisible tools that improve logistics, reduce back-office costs, and make decision-making more efficient.
Conclusion: The bubble may be a necessary step.
Is there a bubble in AI? Everything points to yes, at least in the economic sense. Inflated expectations and minimal returns suggest an inevitable adjustment. But, as in previous bubbles, the real winners will emerge from this correction.
Today, most investments are diluted in unprofitable experiments. Tomorrow, companies that know how to integrate AI with surgical precision—focusing on internal processes, automation, and smart partnerships—will be the ones that capitalize on the real value.
Mientras tanto, conviene recordar una lección básica: no confundir la promesa de la tecnología con su impacto inmediato. La IA no es humo, pero tampoco es magia. Y entre ambos extremos es donde se juega el futuro de esta revolución.
And what do you think, dear friend ?
EN 
ES