By Pablo Revale
Memento
Vol. XIII. 98% of hotels have started using AI. Only 32% are transformed by it. That gap is 126 years old.
June 10, 2026 · 6 Minutes
Adapted from Originally published in Unstoppable Hospitality on LinkedIn.
By Pablo Revale
June 10, 2026 · 6 Minutes
Adapted from Originally published in Unstoppable Hospitality on LinkedIn.
In January, a research report made the rounds: 98% of hotels have started using AI. The industry read it as proof of transformation. Then comes the second number, in the same report. Only 32% say AI is embedded across most of their operations. Everyone is using it. Almost nobody is transformed by it. If that sounds like a contradiction, it isn’t. It’s a pattern. And it’s about 126 years old.
In 1990, the economist Paul David wrote a short paper about why computers weren’t showing up in productivity numbers. Robert Solow had already compressed the problem into one line: you could see computers everywhere except in the productivity statistics. David’s answer: we had seen this before. With the electric dynamo. Edison opened his first central power stations in New York and London in 1881. Two decades later, electric lighting was in 3% of American homes. Electric motors made up less than 5% of factory power. It took four decades for electricity to power half of American industry. And the productivity boom everyone expected? Between 1890 and 1913, productivity growth in the US and Britain slowed down. The two most electrified economies on Earth, decelerating. In 1900 you could have said it about dynamos: everywhere except in the productivity statistics.
Here’s the part that should sound familiar. The first factories to adopt electricity didn’t redesign anything. They bolted electric motors onto the existing system of shafts and belts that steam had powered for a century. Historians call it "group drive." The old transmission stayed in place. The new motors were added on top. Capital went up. Output barely moved. Two systems, one factory, worse math. Now look at your AI stack. A chatbot on top of a PMS from 2003. An AI concierge that emails the front desk, where a human retypes the request into the system. A forecasting model whose output gets pasted into the same spreadsheet as last year. Deloitte’s research on AI returns shows median payback of two to four years and median returns around 10%. Not nothing. Not transformation either. That’s group drive. Wires on top of steam.
Why did factories wait forty years? Why are hotels waiting now? Same reason. Run the numbers from the operator’s chair: the cost of renewal is greater than the cost of integration, and the cost of integration is greater than the cost of the status quo. While that inequality holds, every rational operator picks the cheap end of it. Doing nothing costs the least. Bolting on a pilot is defensible. Rebuilding the operation around the new technology is the most expensive option on the table, while the old plant still works. So nothing structural happens. Until the signs invert. For electricity, three things flipped the inequality. Rates fell hard after 1914. Old factories aged out and their downtown locations became obsolete. And the building boom of the 1920s let companies design factories around the electric motor instead of forcing motors into buildings designed for steam. Then the boom showed up. Roughly half of the acceleration in US manufacturing productivity between 1919 and 1929 traces back to electrification. The gains were never in the motor. They were in the redesign the motor made possible.
What you’re seeing today, the pilots, the chatbots, the 98% adoption headline, is the tip of the iceberg. The mass is underwater: the rebuilt booking flow, the operation designed around guest signals instead of around the PMS, the property where AI isn’t a feature but the architecture. And hospitality has one disadvantage the factories didn’t. Steam engines rusted. Software doesn’t. David warned about this in 1990: information structures don’t depreciate physically. Time alone will never force your redesign. Which means the inversion comes from outside. Google switched on agentic hotel booking in May. Skift reports that travel brands are building AI agents for a consumer that barely exists, since only 2% of leisure travelers will let an AI book for them today. The engineers of 1900 were also designing for a factory that didn’t exist yet. They weren’t wrong. They were early. When guests delegate the booking, when an AI-native competitor runs your market with half your overhead, the status quo stops being the cheap option. The inequality inverts. Standing still becomes the most expensive thing you do. That’s the moment everyone will call sudden. It won’t be sudden. An iceberg flips in seconds, but it melts for years first. The tip was never the story. Watch the waterline.