Where should a hotel start with AI? A practical first 30 days
A practical first month with hotel AI: choose one workflow, involve your team, protect information and test whether the change actually helps.
By
Cindy Hawkins

You don’t need another project that looks great in a meeting and quietly becomes one more thing to manage. If you are considering AI, the first question I would ask is simple: where is your team spending time that could be better spent with guests, clients or each other?
I’ve spent nearly 30 years in hospitality, and I still lead hotel sales today. I know how quickly a day can disappear into reports, follow-up, searching for information and moving the same details between systems. Those responsibilities matter. But they can take a leader away from the work that brought many of us into hospitality in the first place.
My recommendation for the first month is to choose one recurring task, test a small change with the people who do it, and decide whether it genuinely helps. You do not need a property-wide rollout to learn something useful.
Days 1 through 7: Find the work worth improving
Sit with the person doing the task before you choose a tool. Ask them to walk you through the last time they completed it, including the spreadsheet they keep on the side and the information they have to chase down.
Good questions include: Where do you repeatedly start from scratch? What takes several steps because the systems do not connect? What do you have to remember that should be easier to track? Where does a newer team member get stuck?
Look for a task that happens often, has accessible and approved source information, and produces something a person can check. Drafting an internal meeting summary from non-sensitive notes might be a reasonable first test. Letting an AI system change guest reservations without review would be a very different decision.
Pick one task and write down its current starting point: how long it takes, who is involved, what mistakes or rework occur, and what a good result looks like. If your starting point is unclear, you will have a hard time knowing whether the new approach is better.
It is also worth asking whether AI is needed at all. A better template, a shared checklist or a simple reminder may solve the problem with less effort. I want the right solution for the hotel, not an AI label on every process.
Days 8 through 14: Build a small test with clear boundaries
Define exactly what the tool will do and what remains with a person. For example, it may turn approved notes into a draft summary; the department leader still checks the facts, adds context and decides what to share.
Agree on the source information, the expected format and the point where human review happens. Name one person who owns the test. Keep the existing process available so the team can return to it if needed.
Before using real information, confirm which tools and accounts your organization allows and which data may be entered. A personal AI subscription is not permission to upload hotel records. For an early test, use invented examples or carefully de-identified material where practical. Removing a name alone may not make a record anonymous.
NIST’s voluntary AI Risk Management Framework treats trustworthiness as something to consider throughout a system’s design, use and evaluation. For a hotel, that means deciding how a workflow will be checked before it becomes part of everyday work, rather than trying to add safeguards after a problem. [1]
The test should be narrow enough that you can explain it to a busy colleague in a few sentences. If you cannot, simplify it.
Days 15 through 21: Try it during a real working week
A polished demonstration is useful, but the real test is what happens when the person using it has interruptions and other responsibilities.
Work through several representative examples, including an incomplete input or an exception. Ask the team to note where the output is wrong, unclear or harder to use than expected. Include the time spent preparing the input and correcting the result. A tool that produces a draft in seconds may still create more work overall.
Compare the output with the original information. Check names, dates, numbers and any commitments. If the workflow handles a report, reconcile the figures to the source. If it drafts communication, check whether the wording is accurate and appropriate for the relationship.
Invite the person doing the work to disagree. They may see a problem that is invisible from a leadership meeting or a software demonstration. That feedback is part of building a useful system, not resistance to progress.
Days 22 through 30: Decide what deserves to stay
At the end of the month, look at what actually happened. Did the task become easier? Was the output reliable enough for its intended use? Did checking and correcting it erase the time saved? Would the team choose to keep using it?
You might keep the workflow, revise it or stop. All three are fine answers. Stopping an unhelpful process early is better than requiring people to use it because time or money has already been invested.
If you continue, document the basics: the approved tool and source information, the steps, the review responsibilities, the exceptions and whom to contact when something goes wrong. Train a backup person. Decide when to review it again, especially if the source material, tool or business process changes.
What a successful first month looks like
By day 30, I would want you to have one clearly understood problem, a tested approach and enough evidence to make the next decision. That is a stronger foundation than a long list of subscriptions nobody owns.
The larger opportunity is getting your team’s knowledge into processes people can use consistently, while giving leaders more room to lead. Start with one piece of work that matters. Learn from the people closest to it. Build from there.
If you have a task that keeps pulling your team away from guests or sales relationships, tell me how it works today. We can explore whether AI, automation or a simpler process would make it easier.
[1] NIST, AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework
Related service: