Every hotel asking "should we build our own AI tool or buy one off the shelf" is asking the wrong question. The right question is: what does the build-vs-buy decision look like at your specific scale, with your specific data, for the specific use case you have in mind. The answer differs across the four use cases that actually pay back.
Buy when
Build when
The hybrid pattern
In practice, most properties end up with a hybrid. They buy the generic AI features (the PMS-embedded email draft, a SaaS review-response tool, ChatGPT subscriptions for the front office) and build only the workflows that are specific to their data and run at enough volume to justify the engineering cost.
A 240-key resort might buy ChatGPT Team licenses for €25/user/month for all front-office staff (€600/year), buy a review-response tool from Trustyou or ReviewPro (€2k-€8k/year), and build a custom multilingual pre-arrival email workflow on their own MAP system (€40k initial build, €8k/year maintenance). The buy spend is small and immediate; the build spend is justified by 18,000 guest stays per year × €4 of incremental upsell from better pre-arrival comms.
What does not work
Building a custom workflow for a use case where the vendor solution is 95% as good. Building anything before you have ChatGPT or Claude deployed across the team — you cannot specify what to build if you do not know what the off-the-shelf can do. Buying a "fully autonomous agent" before the use case has been proven manually. These are the three most common ways properties waste AI budget.