Eight Questions Every Hotel Leader Should Ask an AI Vendor — Before Signing Anything

AI vendor pitches have become a fixture of the hotel industry calendar. The demonstrations are polished, the ROI projections are confident, and the terminology arrives in quantity: RAG, agentic workflows, MCP connectivity, human-in-the-loop, orchestration layers.

The problem is not that these concepts are meaningless. The problem is that the vocabulary creates an asymmetry — vendors who speak the language fluently, and hotel leaders who are expected to evaluate claims they cannot fully interrogate.

A vendor who struggles with these questions in a sales meeting will struggle more in production. — AI in Hospitality Lexicon, Pertlink 2026

The Eight Questions

1. What AI model or models power this product, and how are they updated?

This question separates genuine AI from AI Washing — the practice of labeling standard rules-based software as ‘AI’ to justify a premium price. Signs of AI washing: the system cannot learn from new data; the vendor cannot explain what model powers the product; outcomes are identical regardless of context.

2. Is our hotel’s data used to train any shared model?

Some vendors train their models on aggregated customer data — meaning your guests’ preferences and your pricing strategies could be contributing to a model that also serves your competitors. Enterprise-grade AI products offer a Closed AI Environment: your data is protected and not used to train shared models. Ask for this in writing.

3. Where is our data stored, and which data protection regulations apply?

Data residency is a governance and compliance question. Do not accept ‘the cloud’ as an answer. Ask for the specific hosting region, the data processor agreements in place, and the regulatory framework under which the vendor operates.

4. Can you show us the output when the AI is wrong — and how we correct it?

Every AI system produces errors. Hallucination — where AI generates confident-sounding but false information — is a known characteristic of language models. In a hotel context, this could mean inventing a rate rule, a facility detail, or a guest commitment. A vendor who can only show you the system working correctly has not prepared for production.

5. What does the approval workflow look like before AI output reaches a guest?

Human-in-the-Loop is one of the most important principles in responsible hospitality AI. For any guest-facing AI output — a review response, a service recovery message, a personalized offer — there must be a named human who can see, edit, and approve before it is sent. Ask the vendor to walk you through exactly how that workflow is designed.

6. What is your incident response process if the AI produces harmful output?

What happens if the AI sends an inappropriate message to a guest? What if it produces a pricing error that undersells a peak weekend? A mature vendor will have a documented incident response process. A vendor who improvises an answer on the spot has not invested in this.

7. How do you handle prompt injection risks in content that the AI reads?

Prompt Injection is an attack vector in which malicious or hidden instructions embedded in external content — a guest email, a review, a supplier proposal — manipulate the AI into violating its rules or leaking data. This is an active area of AI security research. Ask what sanitization and injection protection measures are built into the product.

8. What is included in the SLA — and what is not?

The AI SLA defines what the vendor is contractually committed to deliver. The more important question is often what is excluded: performance guarantees when hotel data quality degrades; support for integration failures caused by third-party systems; liability for AI-generated content that causes commercial or reputational damage.

For hotel leaders, fluency in AI terminology is not a technical ambition. It is a commercial protection. Understanding what RAG means, why hallucination is a structural risk, and why data silos undermine AI performance changes the quality of every vendor conversation and every investment decision.

This article is based on the AI in Hospitality Lexicon (V1.0), published by Pertlink in 2026. Download the full document at www.pertlink.net

Related Document