The AI Dilemma: Should Hotels Wait or Build?

AI is moving so fast it’s really really hard to keep up. One week it’s all about chatbots. The next, it’s local AI agents like OpenClaw executing full workflows on a local device. It’s amazing. But it’s also messy.

This puts hotels in a difficult spot.

Right now, most AI is still probabilistic. That means it gives you a result that is likely to be right, not guaranteed. In marketing, that’s fine. If a chatbot suggests five possible taglines, it’s doing its job. But a PMS can’t guess what rooms are available. An RMS can’t offer a different price every time you refresh the page, depending on which way the wind is blowing. And no one wants their restaurant menu to be reinterpreted every 10 minutes. (Such systems need to be deterministic meaning logical systems that produce identical outputs when given the same inputs.)

At the same time, waiting for the tech to stabilize feels like what hotel industry has done for decades and rarely the best thing.

After the dotcom bust, hotels waited to see if the internet would become a thing. By the time it was obvious, Expedia had already captured much of the distribution stack. We waited on mobile, too. And today most mobile bookings are made through OTAs. When hotels hesitate, others (mostly OTAs) move and pick the low hanging fruit.

Now AI is knocking. And again, the instinct is to wait. Let it mature. Let it become safe. Let someone else figure it out. But here’s the problem: AI will not slow down. And by the time it feels “safe,” others will have already eaten your lunch.

Still, charging into this headfirst with big budgets isn’t the answer either. The landscape is too volatile. Standards don’t exist. There’s a lot of “bricolage” (french for hacking and tinkering things together). One solution that works today might be completely irrelevant in six months. Worse, it might break critical workflows if implemented poorly.

So what’s the right move? (one move is to sign-up for my newsletter here)

There is an idea floating around that maybe, someday, every company (and even individual) will make their own software using AI and that SaaS is dead. It would mean no more off-the-shelf PMS and so forth.

But we are nowhere near that future. I agree it is a possibility and the recent Clawdbot/Openclaw buzz gave us a glimpse of what that could mean. But this is proof-of-concept level. There is still a very long way to go.

Here’s what I think is the right move.

You don’t need to bet the farm. You need a sandbox. A small, focused, skunkworks effort (possibly with a freelancer or AI engineer) working off clean, anonymized hotel data to test ideas. Could you build a better UI on top of your PMS that lets staff request actions to get done and the system executes over multiple clicks and functions? Could you automate parts of group booking administration? Could you prototype whatsapp, email, replies to guests within seconds and adapt it to your style and needs.

Most of it will break. Some of it will be pointless. A few things might be brilliant.

But more importantly, your team will begin to get familiar with what can be done. You’ll learn how to think with AI, not just talk about it. You’ll stop waiting for perfect products and start shaping what you need.

And when the giants do arrive with polished tools and scalable models, the hotels that already have internal experience will be the ones best positioned to integrate, adapt, and move quickly.

Waiting is the easy option. But it rarely works out well in tech, as history has shown. If you’re going to get ahead in this next shift, it won’t come from buying the right product, it’ll come from learning how to build with what’s already here.

Start a skunkworks. That’s your edge.

View source