AI is not a product you buy – it's a capability you build

98% of hoteliers are already using AI in their operations. It's a number that sounds like progress, but Madeline Bushbeck thinks it needs a closer look. 

"If you count someone on the team using ChatGPT to write marketing copy, that's technically AI adoption," says Madeline, Senior Product Manager at Mews. "But it's a long way from having AI that understands how your specific property operates and is actively helping you run it better." 

Bushbeck joined Matt Welle on the Matt Talks Hospitality podcast to discuss what separates hotels that experiment with AI from the ones that actually get something from it. The conversation ranged from data strategy to the limits of automation and offers an argument for building AI capability rather than buying AI products. 

The real gap isn't adoption 

A Mews industry report found that 92% of hoteliers feel optimistic about AI. The enthusiasm is undeniable, but optimism and operational readiness are very different things. 

"There's a huge variance in depth," Madeline says. "Some properties are genuinely sophisticated, using AI across pricing, communications and operations. Then you have properties where the AI conversation starts and ends with: we tried a chatbot on our website and turned it off after a month because it kept saying wrong things." 

The hotels stuck at the chatbot stage often drew the wrong conclusion. Technology has moved fast, and what failed four years ago looks nothing like what's available today. The mistake is treating an old experience as a fixed verdict. 

What Madeline sees most often isn't scepticism about AI. It's hotels running tools that don't know anything about their property. Pricing logic that doesn't reflect how they actually fill rooms. Communications that can't reflect what's actually on offer. AI that was bought off the shelf and bolted on, rather than built into how the operation works. 

"Just because an AI can respond to a guest inquiry doesn't mean it's providing value," she says. "If it's answering incorrectly, or not able to answer and has to surface that to the front desk staff, you're just adding additional work. More noise and frustration than actual value." 

What hotels actually want from AI 

The tech industry often frames automation as something that will remove humans from hospitality processes. But when hoteliers talk about AI, the conversation rarely goes that way. 

"The hotels that are most excited about AI are the ones where the GM says something like: my front desk team is spending two hours a day on operational tasks that have nothing to do with the guests, and I want to give that time back to them." 

The goal isn't headcount reduction. It's restoring the capacity for the work that actually matters in hospitality: being present with guests, building relationships, noticing what someone needs before they ask. That takes time that operational overhead tends to swallow. 

This reframes what good AI looks like for a hotel. It's not about how many steps the technology can remove. It's about how much friction it can take off the people who are supposed to be focused on guests. 

Where automation should stop 

That framing also explains where hoteliers are right to be cautious. 

A pricing algorithm that makes a suboptimal call on a slow Tuesday costs some revenue. That's recoverable. An AI that mishandles a complaint from a guest who drove four hours for their anniversary weekend – that's a relationship that may not come back, and a review that will outlast the incident. 

"Hoteliers understand this intuitively," Madeline says. "And I think they're right to have that caution." 

The Mews approach to closing the gap is a copilot model. AI surfaces the right information, drafts the right message, flags the right moment. A human decides. The efficiency comes from removing friction around decision-making, not from removing the decision-maker. 

Over time, that changes. As staff interact with AI suggestions – accepting some, rejecting others – the system learns. Once the acceptance rate reaches a point where the AI is reliably right, the conversation about fuller automation becomes a different one. But that conversation has to be earned. 

The data problem underneath everything 

Before any of this is possible, there's a more fundamental issue. Most hotels have a data fragmentation problem, and AI can't fix a fragmented operation – it can only accelerate the mess. 

"If your systems aren't talking to each other, your AI certainly won't either," Bushbeck says. "The hotels getting the most out of AI are the ones where reservations, guest profiles, operations and communications are all living in one place, or at the very least, cleanly connected." 

There are two modes of failure: bad data producing bad outcomes faster, and five different AI vendors with no shared context and no central logic. Neither ends well. 

What the semantic layer does 

Even with clean, connected data, there's a layer of knowledge that doesn't exist anywhere in a system. It lives in people's heads. The solution to this is the semantic layer – a contextual knowledge layer that AI agents sit on top of, so they can reason the way anyone at your hotel would reason. 

Without it, any AI making room assignment suggestions is operating blind. Unlike a knowledge base, the semantic layer also learns – updating its understanding of how the property operates as staff interact with it over time. The fix isn't just better AI. It's AI that's grounded in the actual reality of how your property operates. 

How to tell the real thing from the marketing 

Every vendor claims to be AI-powered. The question to ask is: does the AI know anything specific about my property? If the answer is "not really, but you could train it" – that's bolt-on. If it already knows your room types, services, guest history and the institutional knowledge that lives in your staff's heads – without manual setup – that's embedded AI. 

The second test is outcomes. Not features, not demos – what metric is this supposed to move, and is it actually moving it? 

Don't take AI shortcuts

AI is not a product that you buy. It's a capability that you build towards. 

The hotels that come out ahead won't be the ones who found the best vendor. They'll be the ones who understood their own operations well enough to know what to automate, what to protect, and how to make the technology actually work the way their property needs it to.  

It won't happen overnight, but the compounding effect is AI that gets more useful the longer it runs.   

This article is based on an episode of Matt Talks Hospitality. The full episode is available now wherever you watch or listen to your podcasts. 

Watch the episode