The Last Quiet Revolution Was Revenue Management. Agentic AI Is the Next

Decades ago I spent weeks planning for a single New Year's Eve in Goa. I was a young revenue manager, and I began raising our rates early, weeks out, then kept pushing them higher for as long as the booking demand supported it. Then, with the night almost upon us, the demand fell away. The bookings stalled, and every instinct in the building said drop the price and fill the rooms. I held my nerve. And to my surprise, the last minute demand came surging back and paid those higher rates all the way to the very last room. I was fortunate to have a General Manager, Neeraj Chadha, who believed in me, and colleagues who backed the call, and we made history together. The hotel took the highest RevPAR in the country, and the Goa Marriott Resort team was recognised as the best team in the world that year. I tell that story because it captures what revenue management really was in those days. Art and science in the same hands, and anyone who loved the craft could take it somewhere new. The tool we used was tied to the Central Reservations System. We did not even have a proper Revenue Management System back then. We made history anyway, because we believed in the science and knew it could be done. Early and clumsy was never the same as wrong. Saurabh prakash, CEO & founder, phal What we had then was closer to mathematical code than to anything you would now call predictive, a lattice of rules and booking limits rather than a model that learned. The work was slow and unapologetically manual. The logic held all the same. That rough craft is the direct ancestor of the pricing brain inside every serious hotel on earth today. I spent the years that followed watching that back office climb, one decision at a time, into the commercial core of the industry, and in time I led it, revenue, sales, marketing, loyalty and distribution, and eventually operations as well, across a portfolio of some 145 hotels carrying close to a billion pounds a year. The people who learned the discipline early were proved right for the 20 years that followed. I am watching it happen again, the same shape on a far larger scale, a technology taking on a growing share of the decisions that run a hotel. There is a great deal of noise around AI right now, and the number the sceptics reach for is real. Only 11% of organisations have AI agents in production, even as 38% are piloting them Deloitte, Tech Trends 2026 Look closer at why, though, and a familiar pattern appears. The projects that stall are usually agents bolted onto processes that were already broken. As NYU's Nicolas Graf put it this year, AI can free teams from routine work, 'provided the right data foundations and operating model are in place.' That is the whole point. AI is being treated the way we first treated revenue management, left in a corner as a pilot and overridden the moment its advice cuts against instinct, which is exactly when it is worth most. The doubt is the real opportunity. It helps to be honest about where AI actually is. Today's models answer the question you put to them. The frontier has already moved to agents that plan and carry out whole pieces of work on their own. Beyond that sits what the labs are openly building towards, Artificial General Intelligence (AGI), the point at which reasoning, judgement and learning become general capabilities of the machine. Anthropic's Dario Amodei has argued that systems rivalling human experts across many fields could arrive within a few years. Google DeepMind's Demis Hassabis expects it to take a little longer. On the destination they agree, and the destination is what you build for. This is the same shape as revenue management, only steeper. That discipline began as crude code and became the backbone of pricing, taking a larger share of the decisions each year until it sat at the commercial core. AI is early in the same way, rough at the edges and dazzling in flashes, with sound logic underneath. Today it is agentic AI. Tomorrow it is AGI. The groups building with it now will not stay where they are. They will climb the same way a back office of statisticians climbed into the commercial core of this industry. Revenue management was the last quiet revolution that decided who leads in hospitality. Agentic AI is the next. Saurabh prakash CEO & Founder, phal This is why I built PHAL, the first agentic AI embedded operator advisory for hospitality. We build bespoke agentic solutions aimed at the productivity problems that have quietly cost this business for decades. This year's AI-First Hotels report, from NYU SPS and BCG, named the problem exactly. Nearly half of hoteliers report difficulty accessing critical information, and many spend significant time stitching together reports to see a complete picture of the business NYU SPS and BCG, AI-First Hotels, March 2026 That is what PHAL was built for. Every hotel runs on four separate truths, Performance, Profit, Preference and People, and no group has ever held all four in one place with the ability to act on them together. It is the revenue management problem again, one floor up. PHAL OS brings the four into a single decision layer. And we protected the thing that matters most, trust. Our patent pending method is built so that every number can be traced and stood up, so an operator can act on it without arguing with it. Where the work is done properly, the returns are already showing. Early AI deployments have delivered 20% faster room cleaning and preparation, and roughly 50% less food waste within eight months NYU SPS and BCG, AI-First Hotels, March 2026 That is a technology that makes a team count for more, and the operators already using it are proving it. We are only at the beginning. Our first mandate is already moving from proof of concept towards production, a further mandate has a letter of intent is in place, and two more are in the making. The models behind them grow sharper every day. This is the earliest and clumsiest these systems will ever be, and they are already changing how decisions get made. 25 years ago, the people who leaned into a crude, misunderstood discipline rebuilt the commercial future of this industry around it. What sits in front of us now is the same shape, only steeper, and far more powerful, and the doubt sounds exactly the same. General intelligence is the summit this climb is heading for, coming in behind agentic AI. That New Year's Eve, holding to the last room felt like the riskiest thing in the room, and it was the surest call I ever made. I would rather build that future than wait for permission to believe in it, and I am building it with the operators who feel the same.