The problem is not AI, it's business. This is where everything breaks down. A Forbes article in 2026 puts it simply: everyone uses AI on a personal level, but companies don't see results because they haven't redesigned how they work. That's the point. AI is not just another piece of software. Buyer cynicism. Even risks: recent studies show that 78% of users bring their own AI tools to work, and 43% have shared confidential information on them. When the company doesn't integrate... people improvise. And that is more dangerous than not using AI at all. So, what does work? The same studies point to a clear direction: the companies that see results are the ones that stop selling AI as a product... and start operating it as a system. First at home. Then with the client. They become their own 'zero client'. The same pattern. But beyond the anecdotal, the issue is this: the problem with artificial intelligence is not the technology. It's how—or rather, how not—it's integrated into the business. The illusion of pilot projects According to the MIT study 'The GenAI Divide: State of AI in Business 2025,' 95% of attempts to incorporate generative AI do not generate an impact on financial results. It's a matter of consistency. The kill shot The problem is not that AI doesn't work. It's that those who sell it don't use it. It's not technology. I told them that if they gave me access to their platform, I could do exactly the same thing with their tools. It never happened. Months later, at a workshop at Tec de Monterrey, a participant raised his hand. Two and a half years ago, in a boardroom, an executive from a company that sold IBM's artificial intelligence services looked at me incredulously and asked an uncomfortable question: why I talked about OpenAI in my conferences. My answer was simple: I wasn't selling tools, I was explaining a new paradigm. Of easy promises. And that comes at a cost. The real gap What we are seeing is not a technological problem. It's a gap between discourse and operation. Between what is sold... and what is lived. Between understanding the tool... and redesigning work. Because implementing AI is not about installing something. It's about changing processes, decisions, metrics, roles. And that is much harder than doing a demo. It's not a lack of investment, it's a lack of integration. The invisible cost This disconnection already has clear effects. Wasted investment. He asked me this question: — Is what we're seeing scalable in the cloud? I answered carefully: — What you're asking... is exactly what you sell. And this year, an owner of an AI company was direct: they themselves don't use the services they offer. Three distinct moments. They document. And then yes: they sell. It's not magic. It's credibility. The unsolicited advice of today: Before buying or selling AI, ask yourself an uncomfortable question: are you really using this in your operations... or are you just presenting it in PowerPoint? Because if it doesn't live in your day-to-day, it's not going to live in any customer's. And then the inevitable question arises: are you building capabilities... or just narrative? There are few spaces where someone tells you how to do things right, without hype and with real execution. If you are looking for something like that, I leave you this alternative. In promises. And it's not the only one. Another analysis of the same report points out that although 60% of companies evaluate AI tools, only 20% get to a pilot... and only 5% reach real production. That is: most play, but almost no one transforms. McKinsey confirms it: two-thirds of companies are still experimenting, and only 39% report a real impact on EBIT—which is, in short, the real operating profit of a company, what it earns from its business before interest and taxes. It's not a lack of interest. They stay with nice demos. It's a framework for how to use the technology. It's a different way of operating. It's a much more humanistic than technical concept. But many organizations treat it as if it were a plugin. According to various industry analyses, this generates what they call the 'ROI black hole': projects that don't scale because they don't generate value... and they don't generate value because they were never truly integrated into the workflow. It's like buying a Ferrari... to park it. Selling what you don't operate Here comes the uncomfortable part. There is evidence that many companies that sell AI don't even use it internally. Some reports indicate that only 12% have integrated it into their operations, and that up to 95% of their own pilots fail. Yes: they sell transformation... without transforming themselves. This explains something that is increasingly evident: distrust. The report 'The GenAI Divide' from MIT itself collects the perception of executives who have seen dozens of demos, but only one or two are really useful. The market has been filled with 'wrappers'. Team fatigue. It came from AWS. From inflated solutions. They adjust. They measure.
