Why AI in hospitality depends on business context

Almost every AI conversation I have had in hospitality over the last two years has started in the wrong place. It starts with the tool. What can this model do, what can we automate, where do we put a chatbot.

The more useful question is what the system would need to know about your business to be worth listening to.

Generic in, generic out

A general-purpose model asked to write a hotel’s room description will produce something perfectly serviceable and completely interchangeable. That is not a limitation of the model. It is a limitation of the input: nothing in the prompt distinguished this hotel from four hundred others.

The same holds for the more serious applications. Ask a model to suggest where to shift budget and it will give you textbook marketing advice, because it has no idea that your November is driven by a conference centre two streets away, that your spa is capacity-constrained on weekends, or that last year’s discount campaign attracted a guest segment you would rather not repeat.

None of that is in the data. It is in the heads of four people who have worked there for a decade.

Context is an asset that has to be built

This is why we put Insights at the centre of how we work rather than treating it as a reporting service. The understanding layer — what the data says, what the guests do, what the commercial model actually is, why the business works the way it does — is the thing that makes everything downstream better, whether the downstream is a human decision or an automated one.

Building it is unglamorous. It means connecting systems that were never designed to talk. It means writing down the operating context that has only ever been spoken. It means agreeing what a metric means before you put it on a dashboard, and being honest about what your attribution can and cannot support.

That work has always been worth doing. What has changed is the return on it. Business context used to improve the quality of human decisions. It now also determines whether an AI workflow produces something useful or something plausible.

Where it gets interesting

Once the context exists, the applications stop being generic:

• Campaign analysis that knows which revenue streams matter this quarter and which ones are capacity-constrained
• Content generation that has actually read the brand guidelines, the last two years of guest reviews and the tone the team has settled on
• Reporting that flags the anomaly and explains it against what the business already knows about seasonality
• Guest communication that reflects what someone did on their last stay rather than which segment a rule put them in

None of those are technically hard. All of them are impossible without the foundation.

The uncomfortable part

This makes AI readiness much less exciting than it sounds. The businesses that will get the most from it over the next few years are not the ones running the most pilots. They are the ones that did the boring work on data connection, definitions and documented context — often for entirely unrelated reasons.

If you want a single place to start, start there. Not because AI demands it, but because a business that understands itself well enough to explain itself to a machine is a business that was already making better decisions.