Three things every hotel GM should know before taking their first steps with AI

There's no shortage of AI noise, but most of it won’t tell you where to start. Here’s what I’d advise someone tackling it for the first time:

1. Start with the problem, not the technology - The wrong question to start with is "how can we use AI?" The right one is "where does my team spend time on things that shouldn't need a human?” Repetitive guest messaging, manual reporting, rate updates, back-office admin. AI handles all of these well, and the time savings show up fast. Pick one, make it work, then build from there.

2. Context is everything - AI is only as useful as what it knows about your property. A system that doesn't know your rate structure, your guest mix, your team's workflows will produce answers that sound right and aren't. When that happens enough times, your team stops trusting it, and you've lost more than you gained. The question to ask before evaluating anything is: does this tool actually know anything about my hotel, or is it guessing?

3. The goal is more human time, not less - There's a common assumption that the goal of AI is to remove humans from as many steps as possible. Instead, think of it as a copilot. It surfaces the right information, drafts the right message, flags the right moment, and then a human makes the final decision. The efficiency comes from removing the friction around decision-making, not removing the decision-maker. Involve your team early. Explain what technology should and shouldn't do. The hotels that skip this step deploy AI without staff buy-in, create confusion about roles, and end up with low adoption. 

For over 20 years, hotel GMs have been sold the Kool-Aid (occasionally by me) that technology and better use of data would transform their business. First predictive analytics, then big data and now AI. Yet many hotels face fragmented systems, inconsistent data, operational complexity and growing difficulty recruiting staff to sustain legacy service models.

1. Start with one real operational problem. Do not begin with an AI strategy or allow a vendor to dictate the agenda. Choose a genuine pain point for your team: repetitive guest enquiries, slow responses, missed upselling, poor handovers or excessive reporting. Define clearly what success looks like.
2. Don't wait for perfect data. Your data will not suddenly become clean because AI has arrived. Use the best available data for a tightly defined use case, understand its limitations and improve it as you go. Set a realistic MVP and delivery timeline.
3. Decide what stays human. Define where AI can recommend, where it can act and where employees remains accountable. Analysis and repetitive administration are sensible starting points; guest conflict resolution and sensitive decisions still require judgement and empathy.

Takeaway: if AI doesn't improve the experience for both the guest and the team, it is solving the wrong problem.

There's a lot of noise around AI right now, and for most hotel GMs the challenge is figuring out what's actually useful versus what's just hype. What we're seeing at Salesforce is that the starting point doesn't need to be complex, it just needs to be practical.

1. First, make sure your AI is grounded in what you already know about your guest. If a returning guest is still being asked "first time with us?", AI will only scale that problem. Before anything else, focus on connecting and using the data you already have.
2. Second, start internally, not in front of guests. Use AI within the tools your teams already rely on, summarizing shift handovers, surfacing key updates, answering quick operational questions. It's low risk, builds confidence, and keeps anything unproven away from the guest.
3. And third, use AI to give time back to your team. Hospitality is a people business. AI should handle repetitive tasks so staff can focus on the guest. If it's not freeing up time on the floor, it's likely not the right use case.

What we're seeing at Salesforce is that it's not about how much AI you adopt, but how you apply it to real problems.

Having worked inside hotel operations and now alongside hotel teams deploying AI, I have learned that the best first project is usually boring.

1. Start with one real problem, not a broad “AI strategy.” Pick a repetitive, high-volume workflow such as pre-arrival emails or common website questions. Measure response time and staff time before you begin, then run a focused pilot. If success is not defined upfront, every demo can look impressive.
2. AI is only as good as the information behind it. It needs your policies, room details, tone of voice and, where relevant, live rates and availability from your PMS or booking engine. Give someone ownership of keeping that knowledge current. A fast, polished answer is worthless if it is wrong.
3. Automate the predictable and escalate the exceptional. Start with drafts and approvals. Set clear boundaries for complaints, compensation, safety and unusual requests. Involve reception from day one and review real conversations every week.

Do not measure success by how much you automate. Measure faster and more complete answers, fewer repetitive tasks, more booking leads and, above all, more time for the guests in front of your team. The goal is better hospitality.

Every GM should ask -

1. Do you have high quality structured and accurate data in your existing technology (or non-technology based) repositories? If not, then the same GIGO being used to run the business today simply translates into another layer.
2. What information are you not easily able to ascertain today that would make a difference? In relation to a target application of AI, it should provide additional capability or insight that you do not already have.
3. Do you have the internal knowledge base to best technology today, and be able to match that with the next generation of technology; AI? The business of technology is not flippant. If you want a quality outcome with a long-term approach to organisational capability, execution will not be at the end of a prompt. Surround yourself with people who can take your down the road to realisation of the next generation of technology capability.

The more things change, the more things stay the same. AI is just the next layer of technology; on top of all the technology you already have. Not some mystical magical separate world.

Approach technology structurally. That rule has not changed, and it is not going to.

Artificial Intelligence is already reshaping how hotels operate and how guests discover and choose hotels. But with so much noise around AI, it is easy to start with the technology rather than the business problem it is supposed to solve.

If I had to give hotel General Managers just three pieces of advice, they would be these:

1. My first suggestion is that AI should be inclusive: everyone should understand how it can help. In the AI world, that translates to getting everyone an account with whatever chatbot you want to use. (I recommend ChatGPT and Claude because I believe they'll be the leaders.) If you want to get professional instruction from someone who understands training AI better than anyone, buy copies of "Co-intellligence" by Ethan Mollick. Nobody teaches how to understand AI like Mollick.
2. The second point I'd make is that AI will impact everything, but distribution and operations will see the most momentous changes. Agentic travel shopping will result in higher levels of personalization in bookings, and hotels will have to deliver on that personalization. Think about that for a minute. How might AI help you with this impending complexity?
3. Finally, GMs should fight really hard to avoid deploying AI to replace staff. It's short-sighted, and you'll regret it. Hotel distribution is changing to become more personalizable; not every hotel will adapt, but those that do will enjoy significant marketplace advantages. This has implications for operational delivery, as noted above, and the human touch will become more important, not less. We'll need people to deliver the human touch!

1. Fix your data before you buy an agent - An AI is only as good as what it knows. About 80% of what guests want to know lives in your reviews, not your room descriptions — and returning guests shouldn't be treated as a blank form every time. Get your unified guest profiles and review data in order first, or you'll just automate a shallow experience.
2. Start with one skill, one channel — not an overnight overhaul - Think in terms of what AI can do: answer questions, take a booking, upsell, hand over to staff. Switch one on at a time — a web chat agent first, then WhatsApp, then proactive upselling. Phasing keeps risk low and builds your team's trust. Avoid buying long feature lists as siloed tools.
3. Aim for one ongoing conversation - It shouldn't matter who starts it — the hotel with a booking confirmation or marketing email, or the guest with a question or a booking. The end goal is inbound and outbound becoming a single closed loop. That means the outbound system (CDP/CRM) must connect to the inbound one (the agent), so the agent can keep the conversation going without ever losing context.

1. Fix your data first. Compset, segmentation, rate mapping, etc. AI is only as good as the data behind it and most "AI failed" stories are really mapping stories.
2. The bottleneck moves from analysis to decisions. When analysis takes minutes instead of days, the question becomes who decides, how fast, and which old meeting you can delete. It's a change management issue, not a tech one.
3. Start where the loop is fast and the mistakes are cheap. Decide in each case whether AI recommends or decides. Then prioritize areas where you can get feedback quickly.

Here's my take...

1. Educate yourself first - Spend time with the tools yourself before you sit in a vendor meeting. Use them, break them, see where they fail. Take a course if that helps. You cannot judge a demo, or a claim, without your own sense of what the technology can and cannot do.
2. Know your own operation before you buy anything - Most AI projects fail for organisational reasons, not technical ones. Check your numbers, sit with your teams and watch how the work actually happens, then write it down. The real bottleneck is usually in the back office rather than in front of the guest. Be explicit about where you want technology and where you do not, and why. That line is your decision, not a vendor's.
3. Your team decides whether it works, not the technology - If your team does not trust the tool, they will work around it and you will never know. Find the right person inside the team to lead the change and bring others with them. You will need a champion you can trust. Bring the team into the setup, let them ask questions, and make them part of the change. If they need to skill up first, provide the training.

With so much attention on AI, it's easy to focus on the technology and lose sight of what actually drives results. Before making any major investments, I'd keep three things in mind.

1. First, get your foundations in place. AI is only as effective as the data, processes, and systems behind it. If teams don't trust the information they're using today, AI is unlikely to solve that problem. More often, it exposes it.
2. Second, start with a business challenge. Too many organizations begin with the technology rather than the problem they're trying to solve. MIT research examining hundreds of public AI deployments found that only about 5% delivered measurable revenue impact. Success typically comes from applying AI to a clearly defined challenge and measuring results.
3. Third, bring your people with you. Hospitality continues to face staffing pressures, with nearly two-thirds of U.S. hotels reporting shortages. AI can help teams work more efficiently, but adoption depends on trust. Mews research found that hotels with formal AI policies reported nearly twice the trust in AI as those without them.

 The hotels seeing the greatest success with AI are approaching it as an operational and leadership opportunity, not simply a technology investment.

I think we're making the same mistake we make with every technological evolution (again, NOT revolution): we OVERestimate what AI will do in the next six months and UNDERestimate what it will do over the next six years.

Today, I don't think there's a single killer application that every hotel should rush to deploy. But, if I had to pick three trends:

1. Understand how guest behaviour has already changed. More than 60% of searches now end without a click, and that percentage is even higher for top/mid-funnel queries. Start measuring your GEO presence with tools like Semrush or Peec, and invest HEAVILY in content. I recently wrote about this here.
2. Use AI to eliminate reporting. Most hoteliers still spend hours every week pulling data from different dashboards and trying to make sense of it. AI is good at aggregating information, highlighting anomalies and producing meaningful reports in minutes.
3. In order to execute point 2, you need clean data. Every model will eventually become good enough. Whether it's Anthropic or someone we haven't heard of yet is irrelevant. Your competitive advantage will come from having cleaner, richer and better-structured data than your competitors.