AI that doesn't evolve becomes your next legacy problem

What are the questions hotel leaders should be asking now to avoid ending up with a patchwork of AI tools that don’t play nicely together a year from now?

The most important question is also the most boring one: Does this have an open API, and will you actually let me use it?

A surprising number of tools shoot themselves in the foot by building closed, walled platforms. The whole design is to own your data so you can’t easily leave, and that’s exactly how you end up with a patchwork a year from now. If a vendor won’t let your other systems read from and write to theirs, that tells you everything. So ask whether each tool is built to connect or built to trap you. The ones built to connect are the ones that will still be useful when the rest of your stack changes.

Hospitality has been here before with fragmented tech stacks. What’s different about AI, and what lessons from past technology rollouts should hoteliers keep in mind before adding another platform to the mix?

Hospitality has lived through fragmented stacks before: a PMS here, a CRM there, a booking engine bolted on and a small army of hacky integrations and spreadsheets holding it all together. What’s different now is that AI is here to clean that up. All those brittle, custom solutions that used to glue systems together can now be handled simply by agents that move data between systems for you. 

The lesson from past rollouts is that the integration layer is where value quietly leaks out, because every seam is a place where things break. So before you add another platform, ask whether it reduces the number of seams or adds to them. Don't buy another island.

For owners and operators without dedicated AI teams—or even dedicated IT staff—what’s the smartest place to start? How can they experiment without creating more complexity than value?

You don’t need a dedicated AI team, and you don’t even need dedicated IT. What you need is someone who understands the business.

The shift most people miss is this: if you can describe what you want in plain language, you can build it. The barrier used to be technical; now it’s clarity. So the smartest place to start is the process you can explain in a single sentence and feel the pain of every day. Get one agent doing that well, prove the value, then expand.

AI moves fast, but hotel technology investments tend to stick around for years. What separates an AI strategy that's built to evolve from one that’s likely to become tomorrow’s technical debt?

Here’s the test: does it get better automatically when the models get better? That's the whole game. Fine-tuning a model for your specific business feels productive, but it’s mostly a waste of time. The next frontier model will beat your fine-tuned one, and you’ll be left maintaining something the frontier has already lapped. The better use of your time is building a harness—a connection to your whole stack—that simply improves as the models improve. Anything that fights the model’s progress becomes technical debt. Anything that rides it compounds. Build for the second one.

Anything else you’d like to share?

Only that this is the most exciting moment I’ve seen in hospitality tech. For the first time, the technology is moving toward simplicity instead of away from it: fewer systems to wrangle, less glue code, more time back for the people running properties. The operators who win won’t be the ones with the biggest tech teams. They'll be the ones who know their business cold and pick partners that stay open and keep getting better. I like to say AI in hospitality is about making humans more human—less time looking at screens, more time delivering hospitality to the people right in front of you.