Will AI replace hotel staff? What I tell teams
A hospitality leader’s approach to AI and hotel staff: honest conversations, practical training and clear responsibilities for people and tools.
By
Cindy Hawkins

When a hotel leader introduces AI, the team may hear something very different from what the leader intends. You may be thinking about reducing repetitive work. Someone listening may be wondering whether you are trying to reduce the number of people doing it.
That worry deserves a straight answer. I would not tell a team that AI can never affect jobs, because none of us can make that promise. I also would not treat their experience as something a new tool makes less valuable.
When I ran an internal AI survey at a hospitality management company, losing the personal touch with guests came up more often than losing jobs. Accuracy and privacy came up too. Those answers, and the questions I hear in training, shape how I approach adoption. People need more than reassurance. They need to understand what is changing, why it is changing and where their judgment still matters.
Be honest about the change
AI can perform parts of some jobs, and employers will make different decisions about how to use it. The International Labour Organization’s 2025 research describes job transformation as more likely than complete replacement across the occupations it studied. It also makes clear that potential exposure to AI is not the same as actual job loss. That is broad research, not a guarantee for a particular hotel or employee. [1]
My approach is to start with the work a team struggles to keep up with, then ask whether a tool can help them do it better. I want the sales leader spending time with clients, the manager able to coach a new employee, and the team able to find reliable information when they need it.
Those goals should be visible in the project. If leadership says the purpose is to free people for service, it should explain what that means in the working day and how it will evaluate the result. Vague promises will not build much trust.
Show the difference between a task and a responsibility
A system may help assemble a report. A manager still needs to understand an unusual number and decide what to do about it. AI may help draft a sales email. The salesperson still needs to know the client, check the offer and make sure the message is appropriate.
A guest concern is another useful example. A tool might help a team member find the current policy or organize the facts. It should not quietly become the authority for promises, compensation or decisions outside that employee’s role. The person responding needs a clear path to a manager when the situation requires judgment.
These examples make the conversation less abstract. Instead of asking the team to trust “AI,” show them the specific task, its limits and the person responsible for the result.
Let experience shape the system
One of the opportunities I care about most is helping newer hospitality leaders learn from experienced ones. A seasoned general manager often knows what to ask next, which number looks unusual and when a written procedure does not tell the whole story.
Some of that knowledge can be captured in approved reference material, examples and decision guides. AI may help people find and work through that material. It still needs current sources, clear boundaries and an experienced person available when the answer is uncertain.
This is a place to invite the team in. Ask experienced colleagues what a new hire often misses. Ask newer colleagues which instructions are difficult to find or understand. Then build around the real gaps instead of assuming that another login will solve them.
When the people doing the work help define the system, their expertise becomes part of the design. They also have a better chance of recognizing when the system gets something wrong.
Teach people how to question the answer
Training should include an example where the AI is wrong. Show how a convincing response can contain an incorrect date, a made-up detail or an answer drawn from an old procedure. Then practice checking it against the approved source.
NIST’s Generative AI Profile identifies risks including inaccurate generated content and problematic human reliance on AI. Its guidance reinforces the need to evaluate and manage these systems rather than assume a fluent answer is a dependable one. [2]
Give the team a usable review standard. What information must be checked? Which tasks require a second person? What data must never be entered? When should they stop and ask for help?
People also need permission to report a problem without being treated as resistant or incapable. A workflow that looks successful only because nobody feels comfortable challenging it is not ready to rely on.
Make room for different starting points
Some people will experiment immediately. Others will need to see a task demonstrated, try it with support and repeat it a few times. That does not tell you who is a good employee or who can learn.
Offer a small, relevant starting point and time to practice. A front desk team member and a director of sales do not need identical examples. Both need to understand the boundaries and how the tool connects to their work.
Then ask what improved and what became harder. Do not measure adoption only by logins or attendance at a workshop. Look at whether the process is useful, whether the output is accurate and whether people know how to handle exceptions.
The conversation I want leaders to have
My message to a team would be: “Your experience matters in how we build this. We are going to be clear about what the tool does, check whether it helps and keep talking about what needs to change.”
That is a commitment leaders can demonstrate through their actions. It leaves room for honest questions and makes learning part of the work.
Hospitality depends on people noticing, responding and building relationships. As we introduce new tools, I want us to be deliberate about the work we make easier and the human responsibilities we protect. If your team is unsure where to begin, a practical workshop around their own everyday tasks can be a good first conversation.
[1] International Labour Organization and NASK, May 20, 2025: https://www.ilo.org/resource/news/one-four-jobs-risk-being-transformed-genai-new-ilo%E2%80%93nask-global-index-shows
[2] NIST, Generative Artificial Intelligence Profile, July 2024: https://nvlpubs.nist.gov/nistpubs/AI/NIST.AI.600-1.pdf
Related service: