Why We Built Percentage AI (And Where Hospitality Is Going)

From time to time in business, you become aware that the ground beneath your feet is beginning to shift. Not gradually, but decisively. I’ve felt that twice in my career so far. The first time was around 2003, when it became obvious that hotel websites were going to overtake brochures as the primary booking channel. The second time was in late 2023, when I sat in front of an AI tool that wrote a better response to a guest enquiry than the one our team had drafted.

That second moment is why we built Percentage AI. This article is about why we built it the way we did, what we got wrong on the way, and where I think hospitality is genuinely heading over the next five years. It’s a longer piece than usual, because the subject deserves it, and because the AI conversation in our industry has been hijacked by people selling tools rather than thinking about what’s actually changing.

The Real Problem We Were Trying to Solve

For most of the last decade, the central commercial problem in independent hotels has been the same. The OTA channel is convenient for the guest and expensive for the hotel. Direct bookings are cheaper but require more work to capture. The gap between the two is filled, badly, by a stack of tools that don’t quite talk to each other and a website that hasn’t been seriously redesigned in five years.

I’ve spent the bulk of my career trying to close that gap. Ten years inside Triple C Hotels & Resorts in Phuket, building out e-commerce and revenue management. Time before that running IT and revenue at Puravarna Hotels, including the rebrand from Le Meridien Phuket to The Royal Phuket Yacht Club. Years in real estate before that, where I learned that the website is the salesperson, and the salesperson has to be available at 3am in any language.

By the time we founded The Percentage Company in 2018, I’d developed a fairly clear view on what worked. A well-built booking engine, integrated with the right channel manager, paired with smart paid media and a serious CRM, could move a hotel from 12 per cent direct share to 30 per cent direct share within a year. We’ve now done that, on average, across our active client portfolio. The verified number is 30.36 per cent direct revenue share at the 12-month mark, up from a starting average of 13.15 per cent. That’s the work the model was built for.

But by 2023, I started to notice something. The same hotels that had moved from 12 to 30 per cent direct share were plateauing there. The booking engine was working. The ads were working. The CRM was working. But the conversion rate on the website wasn’t moving the way it had been. The guest enquiries were getting longer and more conversational, and our static booking flow couldn’t handle them properly. We were leaving bookings on the table, and we knew it.

That’s the gap Percentage AI was built to close.

Why Existing Solutions Weren’t the Answer

The obvious move, in 2023, would have been to plug in one of the chatbot tools that were already on the market. We looked at all of them. Several of them are well marketed. None of them solved the actual problem.

The chatbot tools were structurally backwards. They were built as overlays on top of existing booking engines, taking guest questions, fetching answers from a knowledge base, and handing the guest back to the booking engine to actually book. That handover was the friction point. Every chatbot deployment we examined had a conversion rate that was lower than the booking engine alone. Putting the bot in front of the booking flow was making things worse, not better.

There was also a deeper problem. The chatbots couldn’t book. They could recommend, suggest, and hand over, but the actual transaction lived elsewhere. So a guest who asked a complex question, got a satisfying answer, and felt ready to book, would be redirected to a separate booking engine where they’d have to re-input everything they’d already discussed. By that point, half of them had given up.

The honest truth is that the chatbot generation of hotel AI was a category error. It tried to add intelligence to a flow that needed to be redesigned, not augmented. We didn’t want to ship another version of that mistake.

So we made a different decision. Rather than build an AI tool that talks to a booking engine, we’d rebuild the booking journey itself, with the AI as the interface and the booking engine and CRM as the connected layers underneath.

That decision took longer to execute than I’d estimated. It also required us to rebuild parts of our own platform that had been working perfectly well for five years. But it’s the decision I’d make again, because the alternative was selling a product I didn’t believe in.

What Percentage AI Actually Is, in Plain Language

Percentage AI is a single conversation that contains the full booking journey. A guest can land on the website, ask any question, and resolve to a booking inside that one conversation. The AI is reading live availability from the booking engine, live history from the CRM, and the property’s actual operational data. It can book, upsell, switch language, recommend packages, and escalate to a human team member when needed. The booking lands in the PMS automatically. The guest record updates. The lifecycle emails get triggered.

It’s not a chatbot. It’s the booking flow, redesigned around how guests actually communicate in 2026, which is conversationally, on mobile, in any language, at any hour.

For owners and general managers, the practical impact is fairly simple. It captures direct bookings that would otherwise have gone to OTAs or been lost entirely. It absorbs the volume of repetitive questions that the reservations team was answering manually. It improves the guest experience in a way that drives repeat bookings and reviews. And it gives the property a system that doesn’t require constant human supervision to keep running.

What I Got Wrong on the Way

It would be dishonest to write a thought leadership article about building a product without admitting where the thinking was wrong on the first attempt. So I’ll be direct about a few things.

First, I underestimated how much the underlying data architecture would matter. We started by trying to build the AI on top of our existing stack and ran into the same problem the chatbot vendors were running into. The AI was always slightly out of sync with what the booking engine and CRM were doing. We had to rebuild the data architecture before the AI could work properly. That took months we hadn’t planned for.

Second, I overestimated how much guests would want a “concierge experience” and underestimated how much they wanted reliable answers to boring questions. The early prototypes were clever. They could recommend wine pairings and arrange yacht charters. The problem was that 90 per cent of real guest questions were about whether breakfast was included, what the cancellation policy was, and whether the pool was heated. A clever AI that didn’t reliably answer the boring questions wasn’t useful. We had to rebuild around the boring questions first, then layer the cleverness on top.

Third, I assumed hotels would need a long onboarding process to adopt this, when in practice the smart properties wanted to move quickly. Our full-service implementation now runs at 4 to 6 weeks from signed agreement to fully live, including tech stack setup, booking engine deployment, and AI launch. That’s faster than I thought it should be when we started, and it’s working.

The lesson, if there is one, is that building hospitality technology requires fewer assumptions about what hotels “should” want and more attention to what they actually do every day. The clients teach you the product if you’re willing to listen.

Where Hospitality Is Going

I want to be careful here, because every article about the future of hospitality eventually descends into either AI hype or AI panic, and neither is useful. Let me try to be specific about what I think is changing and what isn’t.

What’s changing is the booking journey. The traditional hotel website, with its homepage, room pages, and booking engine widget, is going to look as dated by 2030 as a 2005 website looks now. Guests already expect to be able to ask questions and get answers. Within two or three years, that expectation will harden into a requirement. Properties that haven’t redesigned their booking journey around conversation will lose direct bookings to those that have, and will be pushed deeper into OTA dependency to compensate.

What’s also changing is the role of the reservations team. The team that spends its day answering repetitive questions is going to look as inefficient by 2028 as a hotel that doesn’t accept credit cards looks now. The reservations role will shift towards higher-value guest interactions, complex bookings, and revenue recovery. That’s a healthier role and probably a more interesting one, but it requires the AI to do the volume work reliably.

What isn’t changing, at least not in the timeframe most articles imagine, is the central importance of the human element in hospitality. Guests will still want to be greeted by a real person, served by a real person, and remembered by a real person. The AI is there to handle the things that don’t need a human, so that the humans can focus on the things that do. Any vendor who tries to sell a hotel an AI that “replaces” the front desk is selling something that won’t survive contact with reality.

Underneath all of this, the commercial logic is the same logic that’s been true since 2003. Hotels that own their guest journey win. Hotels that depend on intermediaries lose, slowly. The AI is the most powerful tool we’ve yet had for owning the guest journey, but it’s a tool, not a transformation. The hotels that use it best will be the ones that already understand what they’re trying to do commercially, and use the AI to do it faster and at lower cost.

What I’d Tell a Hotel Owner Reading This

If you’re an owner or GM reading this, and you’re trying to work out what to do about AI for your property, I’d give you three pieces of advice based on what I’ve seen.

The first is to be sceptical of any AI tool that doesn’t book. If the AI sits in front of your booking engine but can’t actually complete a transaction, you’re buying a chatbot, and chatbots don’t move the commercial needle. The conversion lift comes from the integrated booking flow, not from the chat interface alone.

The second is to focus on what your guests actually ask. The boring 80 per cent of questions, breakfast, parking, cancellation, transfers, are where direct bookings get lost. If the AI handles those reliably and at speed, the commercial impact will follow. The clever stuff comes later.

The third is to think about it as a system, not a feature. The booking engine, CRM, and AI need to be connected, not stitched together. If you’re being sold an AI tool that “integrates with” your existing stack, ask hard questions about what that integration actually means in practice. The gaps between systems are where bookings die.

The Closing Argument

We built Percentage AI because the existing options were going to disappoint our clients, and we couldn’t sell them in good conscience. We built it the way we did, with the booking engine, CRM, and AI as a single connected system, because that’s the only way the commercial numbers hold up over a 12-month period rather than for the first month after launch.

The bigger picture is that hospitality is moving into a phase where the properties that get the technology right will pull away from the ones that don’t. The gap won’t be visible immediately. It’ll be visible in three years, when the properties with integrated AI booking journeys are running at 35 to 50 per cent direct booking share with leaner operational teams, and the properties that stuck with their 2018 stack are still paying 20 per cent OTA commission on most of their bookings.

I’d rather our clients were on the right side of that gap. That’s what Percentage AI was built for, and that’s why we’re putting it in front of properties that take their direct bookings seriously.

If you’d like to talk about what this would look like for your property, you know where to find us.