The real problem isn't a lack of data. It's a lack of connection.
We're at a point in time where most people use AI in some capacity daily. Yet, only 43% of professionals actually trust the raw outputs of AI tools (Source: McKinsey).
When we apply that to the hotel industry, we're seeing hoteliers are asking AI for recommendations, and AI is communicating likely flawed answers without clear reasoning because it doesn't have the full context, in turn degrading trust further.
People are asking AI to replace their thinking, when its real power is to connect their data.
And the distinction matters enormously — because the same technology that produces a confident, reasonless recommendation is also capable of surfacing the kind of insight that would take a manual process months to find, if it even manages to.
I've been on my own journey with AI over the past two years — from skepticism to running AI agents across our engineering, analytics, and reporting workflows daily. What I've learned is that most of the hype, and most of the frustration, comes from the same misunderstanding: people are asking AI to replace their thinking, when its real power is to connect their data.
When we use AI effectively to allow systems to talk to each other we unlock incredible potential. The puzzle that's been impossible to solve suddenly comes together and you understand more about your hotel than you thought possible.
Revenue managers today sit at the center of a paradox. More data than ever before. Fewer clear answers. Each system is speaking its own language — the PMS knows guest history but can't explain booking patterns; rate shopping shows competitor prices but can't suggest strategy; the market calendar shows events but can't connect them to demand impact; and in many cases the revenue system prices rooms without context from any of the others.
Without intervention, it's impossible for them to understand each other.
The result: revenue managers spend the majority of their day as translators, manually carrying information between systems that were never designed to speak to each other. We've observed a majority of revenue managers still spending over half their day on data compilation and report generation.
That is the real complexity problem. And it is not solved by better systems. It is solved by connecting the ones you already have.
Here is where most hotels go wrong. They try to train AI rather than connect it. They build new data pipelines instead of working with what's already in place. They spend months teaching AI things their team already knows about their property and their market. They lose institutional knowledge in the process and end up with a system that starts from zero every time.
The smarter approach is to use AI as connective tissue — a translation layer that enables every platform you already run to share context and work together. One AI layer bridges PMS, revenue management, bookings, analytics, guest services, and operations. Data flows freely. Patterns that were invisible across siloed reports surface automatically.
Each platform speaks its own language until a translation layer lets them share context.
When systems can finally talk to each other, intelligence emerges. The "why" behind RevPAR movements, the connection between a festival weekend, booking pace, competitive rate shifts, and demand patterns. They all stop being a puzzle you reconstruct manually and become a question you simply ask and get answers to.
Build on existing systems and workflows. Preserve your team's institutional knowledge. The goal is not to rebuild your operation around AI — it's to give your existing operation a layer that makes it significantly more capable.
Hotel operators drive the analysis. Not IT. Not data scientists. AI provides the processing power; your people provide the context that makes it meaningful. Hotel expertise — not technical skill — should be directing every query. The teams getting the best results from AI are not the most technically sophisticated. They are the ones who know their business best and have learned to direct AI with that knowledge.
Skip the transformation initiative. Use the data you already have. The right AI implementations produce results in days, not months — and that early proof of value is what builds the confidence to scale.
So how do you get there? AI adoption works best as a deliberate climb. Each stage builds the trust and infrastructure needed to reach the next level safely.
AI adoption works best as a deliberate climb — click a level below to see how risk, oversight, and payoff change.
Ask questions, explore patterns, validate AI outputs against what you already know. This is where familiarity is built and where most teams should start.
Ask your PMS why pickup was flat last Tuesday.
Most teams should start at Level 1 and earn trust before advancing.
The temptation is to skip to Level 4 because the promise is compelling. The teams that succeed treat this as a marathon, not a race.
This 4 level process is just as much about gradual improvement as it is building trust in AI within your team. We've followed this process at Duetto with promising results: since January 2026 to the close of Q1, we've increased AI-use by our R&D engineers by 47%, a signal of trust as well as successful implementation.
Forget the transformation roadmap. These four experiments are designed for immediate action — each one low risk, each one capable of delivering something concrete.
No SQL. No data science. Take the data you are already working with, give it context, and ask AI a plain-language question about your property. The goal is not a perfect answer — it's to get the AI to tell you something about your own data that you didn't already know.
Connect your PMS, comp-sets, and market calendar. Ask: "Why did RevPAR drop during the festival weekend?" This is where the connective tissue value becomes visible — the connections that siloed reports have always missed surface in a single response.
Let AI identify guest behavior patterns automatically. Forget manual segmentation. The aim is intelligence that surfaces without you having to build the query. Low effort. Surprisingly high signal.
Revenue asks about shoulder periods. Operations asks about staffing. Finance asks about margin contribution. Get rid of templated responses, and ask AI to analyze the data and provide a different answer to each person. No need for generic reports that nobody reads in full anyway.
I've seen the same three failure patterns repeat across organizations, and recognizing them early is the difference between a costly lesson and a genuine result.
The trap is attempting to do too much with AI at once. The fix is mastering one high-value use case before expanding. Breadth without depth produces noise, not insight.
You're going to run into problems if you use AI you can't explain or interrogate. Any AI that makes a confident recommendation without visible reasoning is dangerous — because you have no way to know when it is wrong. The fix is keeping humans in the loop and demanding explainability. AI should augment your judgment, not replace it.
Don't buy features without a workflow plan. Define the outcome first, then select the tool. Acquiring AI capability is easy. Using it to produce a specific result that improves a specific decision is harder, and requires clarity about what you actually need before you spend.
A traditional revenue manager's day looks something like this: 60% data compilation, 25% report generation, 15% strategic analysis. An AI-enabled day inverts the ratio: 15% data validation, 25% insight exploration, 60% strategic work. The same person, the same expertise — doing the work that actually changes performance outcomes instead of the work that produces the inputs to that work.
The same person, the same expertise — strategic work grows from 15% to 60% of the day.
The turning point for me — and I've heard this from others too — wasn't a product demo. It was the first time AI answered a question I hadn't thought to ask. That's the moment the technology stops feeling like a tool and starts feeling like a genuine capability shift.
The question every revenue manager, every GM, and every commercial team is asking right now is not whether AI will change hotel operations. We all know it will.
The question is whether you are building the foundations today that will make you a leader of that change, or whether you are watching from a position you will later have to recover from.
Start with one experiment this week. Build trust before you scale. Use the data you already have. Making meaningful improvements with AI doesn't begin with a transformation initiative. It begins with a single conversation between you and your data.
Dive even further into your hotel's overall performance, beyond AI, with performance engineering — a new way of operating and thinking for hoteliers. This approach is the natural build on using AI to connect your data, focused on aligning your teams and workflows to a profit dataset.