Not All AI is Created Equal (And You’re Probably Reaching for the Wrong Kind)

One of the most clarifying moments at MURTEC 2026 came when Sol Rashidi drew a line that most AI vendors have a financial incentive to blur. Not all AI is the same. Not all AI requires the same organizational readiness. And the hospitality industry, in its rush to deploy, is systematically reaching for the most complex category of AI while still building the foundational infrastructure the simplest category requires.

Rashidi has earned the standing to make that observation plainly. More than 200 AI deployments across industries, 44 still running. The rest are case studies in exactly the mismatch she described at MURTEC. Her message to a room full of hospitality technology leaders was direct: before you decide what AI to buy, you need to understand what kind of AI you are actually capable of deploying.

Those are not the same question, and conflating them is expensive.

The Three-Tier Framework Every Hospitality Operator Needs to Understand

Rashidi’s framework distinguishes three fundamentally different categories of AI, each carrying distinct requirements for infrastructure, governance, talent, and organizational maturity. Understanding where your organization sits relative to each tier is the prerequisite to every other AI decision.

The progression matters. Applied AI is the foundation. It requires clean data, system integration, and disciplined workflow design, but it keeps humans in the loop and produces measurable, defensible results. Generative AI adds creative and content capability but introduces data leakage risk that hospitality operators, sitting on rich guest behavioral data, cannot treat casually. Agentic AI operates autonomously, making decisions at machine speed without human validation at each step.

Each tier up the stack multiplies the governance burden, the security requirements, and the organizational change management challenge. Most hospitality operators are not ready for the tier they are reaching for.

The Tricycle and the Ducati

Rashidi’s analogy landed hard in Las Vegas, and it deserves to be repeated in every hospitality boardroom evaluating an AI vendor pitch right now.

“If you’re still learning to ride a tricycle, try not to get on a Ducati so quickly.“

For a full-service hotel or restaurant with fragmented data across a property management system, a CRM, food and beverage platforms, a spa booking system, and a loyalty program, the gap between current infrastructure maturity and agentic AI requirements is not a minor technical challenge. It is a fundamental organizational readiness problem.

The vendor community is not helping. Agentic AI is the category generating the most excitement, the most investment, and the most aggressive sales activity in hospitality technology right now. Autonomous booking agents, self-optimizing dynamic pricing engines, predictive maintenance systems that dispatch technicians before guests notice a problem. The pitch is compelling. The problem is that organizations still building the data pipelines Applied AI requires are being sold Agentic AI deployments they are not structurally prepared to sustain.

The 74 to 88% failure rate Rashidi cited in her MURTEC keynote (see Post 1) is not a coincidence. It is the predictable consequence of this mismatch, repeated across hundreds of enterprises simultaneously.

Where the Hospitality Industry Actually Is

The honest assessment of where most hospitality operators sit in this framework is uncomfortable, but necessary.

Applied AI demands clean, connected data. A single-property independent hotel with a modern PMS and a reasonably integrated CRM may genuinely be Applied AI ready. A multi-property management company running six different PMS platforms across its portfolio, with loyalty data siloed in a corporate system that syncs nightly, is not.

Generative AI requires disciplined data governance and active security controls. Guest profiles, payment behavioral data, pricing strategy, and competitive rate intelligence are exactly the data categories that generative AI systems are most likely to surface, combine, and potentially expose through poorly governed prompting or unsanctioned usage. Most hospitality operators cannot yet account for all the ways their data is being accessed, by whom, and whether that access is sanctioned.

Agentic AI requires all of the above plus the organizational trust and governance infrastructure to allow machines to make consequential decisions autonomously. In an industry where a single guest interaction can generate a TripAdvisor review, a social media post, or a loyalty defection, the consequences of an autonomous agent making the wrong call are brand-level, not just operational.

The Procurement Use Case: What Good Looks Like

Rashidi singled out procurement as her favorite AI use case across all industries, and the reasoning applies directly to hospitality.

Procurement is overworked, understaffed, managing significant contract volume, document-heavy, and universally frustrated with slow turnaround times. It is also a back-office function where AI errors carry operational consequences but not brand consequences. A contract review that takes an extra day to correct does not generate a guest complaint.

For multi-property hospitality operators, the procurement profile fits precisely. Contract management across ownership groups and management agreements, vendor negotiations across hundreds of food and beverage SKUs, compliance documentation for brand standards, and capital expenditure approvals represent exactly the kind of high-volume, document-intensive workflow where Applied AI delivers measurable, defensible ROI.

“If your goal is to cut a piece of paper in half, why would you use a chainsaw if scissors can do the trick?”

Hospitality operators chasing AI-powered revenue management when their labor scheduling is still done in spreadsheets have the sequence wrong. Fix the sequence before you scale the technology.

The Readiness Assessment Every Operator Should Run

Before the next vendor meeting, before the next board presentation on AI strategy, Rashidi’s framework demands an honest answer to five infrastructure questions:

How much data do we have, and where does it live? If the answer requires a multi-department audit to produce, Applied AI readiness is aspirational, not actual.

Is our data usage sanctioned and governed? If team members are using consumer generative AI tools to process guest data without an organizational policy governing that usage, Generative AI risk is already present whether or not a formal deployment exists.

Who has access to our data, and who should not? The absence of a clear answer to this question is not a data governance gap. It is an AI security liability

Have we redesigned our workflows for AI, or are we automating existing ones? Automating a broken workflow produces a faster broken workflow. The sequence is optimization first, automation second

Do we have humans positioned to validate AI outputs upstream, not downstream? Rashidi’s hospital case study, which Post 4 in this series will examine in detail, demonstrated precisely what happens when humans are positioned at the end of an AI workflow instead of the beginning. The consequences in hospitality are the same, with the added dimension of brand and guest relationship risk

The Bottom Line

The hospitality industry does not have an AI ambition problem. It has an AI sequencing problem.

The organizations that will build durable AI capability in this industry are not the ones buying the most advanced technology available. They are the ones that accurately assess where they sit in Rashidi’s three-tier framework, deploy at the tier their infrastructure actually supports, build the governance and data foundations the next tier requires, and then move up the stack deliberately.

Rashidi closed this section of her MURTEC keynote with a framing that should anchor every technology strategy conversation in hospitality for the next 18 months: “Technology scales efficiencies, but in relationship-driven industries, relationships scale opportunities.”

The tricycle is not a consolation prize. It is the vehicle that builds the skills to eventually ride the Ducati safely. The operators who understand that will still be running their AI deployments when the operators who skipped the sequence are explaining to their boards why the POC never made it to production.

Up Next in the Series:

This was Post 2. Post 3 examines the critical difference between “Using AI” and “Doing AI”, and why most hospitality players are investing in the former while expecting the results of the latter.

IHL Group covers retail and hospitality technology markets globally. For more information on our research, visit https://www.ihlservices.com. Sol Rashidi keynoted MURTEC 2026 in Las Vegas. All data and frameworks cited in this post are attributed directly to her presentation.