What does AI in hotels mean?
AI in hotels includes statistical forecasting, machine learning, optimization, rules-based decision aids, and generative systems that produce predictions, recommendations, allocations, or actions for hotel work. Hotel Decision Lab studies systems that inform or execute managerial decisions. The term is broader than generative AI and does not mean that every hotel technology product is intelligent.
The core Hotel Decision Lab question
How should decision authority be divided between hotel managers and intelligent systems?
This question is narrower than “How can hotels use AI?” It asks how a hotel should allocate authority when a system and a manager may hold different information, detect different errors, face different incentives, or control different parts of a decision.
Where AI enters hotel operating decisions
Prediction
What is likely to happen?
Examples include occupancy, demand, workload, labor need, no-shows, service volume, maintenance risk, or the probability of a disruption.
Resource allocation
How should limited resources be assigned?
Examples include staffing levels, schedules, task assignments, room inventory, time, attention, or operating capacity.
Real-time personnel action
What should happen now?
Examples include calling in an employee, sending someone home, reallocating work, pausing a process, or escalating an exception.
Decision quality
What outcome should the hotel evaluate?
Accuracy is only one dimension. Labor cost, service, employee burden, guest consequences, fairness, reversibility, and accountability may also matter.
A model may be accurate at prediction while the organization still makes a poor allocation or personnel decision. Separating these stages helps identify where information, error, authority, or accountability failed.
Why “human in the loop” is not enough
A human approval step describes that a person is present. It does not show that the reviewer holds useful additional information, can recognize when the system is wrong, or has real authority to intervene.
Hotel Decision Lab proposition
Informational advantage
The manager contributes material, timely information the system does not possess or represent adequately.
Hotel Decision Lab proposition
Diagnostic competence
The reviewer can recognize consequential system errors rather than reacting mainly to discomfort, habit, or confidence.
Hotel Decision Lab proposition
Governance alignment
The reviewer has meaningful authority, adequate time, aligned incentives, and accountability for the intervention.
What remains unknown
These are not validated hotel rules.
The three conditions are proposed constructs for evaluation and future testing. Current evidence does not establish them as necessary-and-sufficient conditions in hotel labor operations.
Read the full argument in When Hotel Managers Override the Algorithm.
What current evidence supports
Transferred evidence
Calibrated reliance is the goal.
Responsible use is not maximum trust or maximum skepticism. Reliance should vary with comparative capability, uncertainty, task conditions, and consequences.
Transferred evidence
Aversion and appreciation both occur.
People may avoid algorithms after observing error, yet in other settings prefer algorithmic judgment. A single universal attitude toward AI is not supported.
Transferred evidence
Control can improve uptake without improving accuracy.
Modification rights can increase willingness to use an algorithm. Other work shows that adding human intervention can also reduce decision accuracy.
Transferred evidence
Explanation is not automatically diagnostic.
An explanation may affect trust or acceptance without helping a reviewer determine whether the recommendation is correct or whether the human-AI team performs better.
These patterns draw substantially from adjacent domains. They motivate hotel research; they do not prove that the same effect size or boundary condition will appear in every hotel decision.
What hotel-specific evidence does and does not establish
Hotel-specific evidence
AI-supported decisions are operationally relevant.
Hotel research documents occupancy-prediction applications, hospitality employee responses to AI, and hoteliers’ perceptions of automated decision-making across hotel functions.
What the evidence does not establish
Manager intervention has not been tied cleanly to downstream outcomes.
The current hotel-specific corpus does not establish when a manager’s acceptance, modification, override, pause, or escalation improves labor cost, service, employee, guest, fairness, or accountability outcomes.
Measures of perception, intention, trust, or adoption are useful but are not substitutes for decision quality. A hotel can increase system use while also increasing overreliance, superficial approval, or poorly justified intervention.
Decision rights: four proposed operating modes
At design time, a hotel can assign different levels of authority to a system and a manager. The following modes are a proposed vocabulary, not a validated framework:
Proposed mode 01
Human-led
The manager makes the decision and may consult system information.
Proposed mode 02
AI-recommended
The system proposes an action and the manager decides whether and how to act.
Proposed mode 03
AI-generated with required review
The system produces the default decision, but a defined reviewer must assess it before execution.
Proposed mode 04
Automated with human monitoring
The system acts unless a threshold, exception, or later review triggers intervention.
The appropriate mode may depend on demonstrated system capability, local information, consequence severity, error detectability, reversibility, frequency, and the organization’s ability to monitor outcomes. Those relationships remain research propositions rather than settled hotel rules.
Run-time responses need more than accept or override
Even after design-time authority is assigned, the reviewer needs a defined response vocabulary:
- Accept: execute the recommendation as given.
- Modify within bounds: adjust only within a pre-defined range or set of fields.
- Pause: stop execution while awaiting information, a threshold, or a higher-authority decision.
- Investigate: collect additional evidence about the model, input, local condition, or consequence.
- Override: replace the recommendation with a different action and record the reason.
- Escalate: move the decision to a person or process with different authority or expertise.
This distinction matters because “human review” can otherwise conceal very different authority structures.
The current research program
Hotel Decision Lab’s first publication examines manager overrides and human review. The editorial program then develops local knowledge, explainable AI, appropriate reliance, and hotel AI decision rights as connected parts of the same research question.
- When Hotel Managers Override the Algorithm
- When Local Knowledge Beats the Forecast
- AI in Hotel Operations Research Agenda
- All Hotel Decision Lab articles
What remains unknown
- When does manager-only information materially improve an AI-supported hotel labor decision?
- When do hotel manager overrides correct a model, and when do they introduce bias, noise, optimism, or organizational pressure?
- Which decisions should permit bounded modification, unrestricted override, pause, escalation, or full automation?
- Can uncertainty displays or explanations improve error detection rather than merely change confidence?
- How should a hotel define decision quality when labor cost, service, employee burden, guest consequences, fairness, and reversibility conflict?
- Who should bear accountability when prediction, allocation, approval, and execution are distributed across a system, vendor, manager, and organization?
Method note
This page is based on Hotel Decision Lab’s structured integrative review and current research-gap synthesis. It separates hotel-specific evidence from findings transferred from forecasting, human-factors, organizational, consumer, and other decision settings. Read the full research methodology and evidence standards.
Selected references
Athey, S. C., Bryan, K. A., & Gans, J. S. (2020). The allocation of decision authority to human and artificial intelligence. AEA Papers and Proceedings, 110, 80–84. https://doi.org/10.1257/pandp.20201034
Bansal, G., Wu, T., Zhou, J., Fok, R., Nushi, B., Kamar, E., Ribeiro, M. T., & Weld, D. S. (2021). Does the whole exceed its parts? The effect of AI explanations on complementary team performance. Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, Article 81, 1–16. https://doi.org/10.1145/3411764.3445717
Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General, 144(1), 114–126. https://doi.org/10.1037/xge0000033
Dietvorst, B. J., Simmons, J. P., & Massey, C. (2018). Overcoming algorithm aversion: People will use imperfect algorithms if they can even slightly modify them. Management Science, 64(3), 1155–1170. https://doi.org/10.1287/mnsc.2016.2643
Ivanov, S., & Webster, C. (2024). Automated decision-making: Hoteliers’ perceptions. Technology in Society, 76, 102430. https://doi.org/10.1016/j.techsoc.2023.102430
Kozlovskis, K., Liu, Y., Lace, N., & Meng, Y. (2023). Application of machine learning algorithms to predict hotel occupancy. Journal of Business Economics and Management, 24(3), 594–613. https://doi.org/10.3846/jbem.2023.19775
Lee, J. D., & See, K. A. (2004). Trust in automation: Designing for appropriate reliance. Human Factors, 46(1), 50–80. https://doi.org/10.1518/hfes.46.1.50_30392
Logg, J. M., Minson, J. A., & Moore, D. A. (2019). Algorithm appreciation: People prefer algorithmic to human judgment. Organizational Behavior and Human Decision Processes, 151, 90–103. https://doi.org/10.1016/j.obhdp.2018.12.005
Sele, D., & Chugunova, M. (2024). Putting a human in the loop: Increasing uptake, but decreasing accuracy of automated decision-making. PLOS ONE, 19(2), e0298037. https://doi.org/10.1371/journal.pone.0298037