AI in Hospitality, Part 1 of 15
A consultant comes to the property and talks about AI assistants, automated workflows, maybe a digital concierge. Everyone nods. Three months later, either nothing is running at all — or a single chatbot that doesn't really help anyone.
Usually there's no big catastrophe behind this. Just a gap nobody looked at beforehand.
What are data silos in a hotel operation?
Data silos arise when guest information is spread across multiple systems that are not connected to each other — for example PMS, point-of-sale system, accounting, HR software and CRM. Each system stores only a fragment of the information about the same guest, without the other systems knowing about it.
Five systems, five truths
A typical operation works with a PMS for room management, a point-of-sale system for restaurant and bar, accounting software, an HR solution for shift schedules and payroll, plus a CRM or an Excel list for regular guests. Five systems. They rarely talk to each other.
That's not a problem in itself. The problem arises because each system only knows a fragment of the same guest. The PMS knows that Ms Meier prefers a room with a balcony. The CRM knows this is her fourth booking. The kitchen knows she eats gluten-free. None of the three knows what the other two know.
This is not an isolated case: according to the 2026 Hotel PMS Impact Study by HotelTechReport, 45 percent of hoteliers surveyed name faster integrations and open interfaces as the most urgent development need for their PMS. And according to the HEDNA State of Distribution Report, 67 percent of independent hotels describe dealing with separated systems as one of their biggest operational challenges.
Why this becomes a problem for AI
An AI assistant that pre-sorts emails, answers inquiries or forecasts occupancy only sees the data it's given. If this data is scattered across five systems, it sees at best half the picture. It answers an email correctly and doesn't know that the same guest just had a complaint in their room. It suggests a room that, according to housekeeping, isn't ready yet.
The obvious reaction: "The AI doesn't work." In fact, it did exactly what was possible with the available data. The actual problem sits one level deeper.
First the tool, then the disappointment
A common pattern: a property hears about a promising AI tool, introduces it — and is surprised by the lack of impact. Rarely is this the tool's fault. Usually it's a matter of sequence.
Before an assistant, an automated workflow or a standalone assistant can work meaningfully, answers are needed: What data exists? Where is it located? Who is allowed to access what? And, not to be underestimated especially with data protection: What is legally permissible in the first place? This is less spectacular than a tool demo. But it's exactly this work that later determines success or frustration.
An exercise for your own operation
Trace the path of a single guest through your systems — from the booking inquiry to the invoice. At how many points does someone transfer information by hand from one system to another? And how often does a preference, a note or a special arrangement get lost in the process?
Every one of these points is an interface. If it stays open, any later AI application will hit a dead end there. A structured stock-take therefore doesn't start with the question of which tool, but with the question of the foundation.
What comes next in the series
The upcoming articles will cover AI in hospitality — not only in marketing, but also in staff planning, accounting, front desk, revenue management and internal communication. What connects all these articles: a clean, consolidated data foundation determines whether AI can have any effect in a given area at all.
The next part looks at how to connect systems in a meaningful way, without immediately overhauling the entire IT landscape.
Further reading
- Agentic AI: Why hotels now need to rethink their booking systems
- Why don't hotels have their own AI budget — and what does it really cost?
- AI in the hotel: start now or wait for the next generation?
Sources
- HotelTechReport (2026): 2026 Hotel PMS Report: What Elite Operators Want, Use, and Plan Next
- HEDNA State of Distribution Report, cited in Cloudbeds (2026): PMS Integrations: What They Mean for Hotels in 2026
Frequently asked questions
Why do AI projects in hotels often fail?
Usually not because of the chosen tool, but because the underlying data is spread across multiple separate systems (PMS, CRM, accounting, HR) and nobody consolidates it. An AI assistant can only work with the data it's allowed to see.
What is the first step before introducing AI in a hotel?
A structured stock-take of systems, data, team and processes — before a tool is selected. Tourismusconsult offers the AI status assessment for this purpose, which leads to a prioritised roadmap with three time horizons.
Which systems in hotels are typically separated from each other?
Most commonly: property management system (PMS), point-of-sale system, accounting software, an HR solution for staff planning, and CRM for guest relations.
Tourismusconsult supports hotel operations with the AI status assessment: a structured analysis of systems, data, team and processes that leads to a prioritised roadmap — with quick wins, medium-term and strategic measures.
