Start with the decision, not the feature
Adding a model to a workflow does not automatically make the workflow better. A more practical starting point is a question a manager already has to answer: where should the team investigate demand changes, which maintenance requests need attention, or how should a pricing recommendation be reviewed? The decision defines the information, timing, user, and acceptable level of uncertainty the system must support.
Demand analysis, maintenance triage, and pricing recommendations may be good candidates for intelligent support, but each needs a definition of success and a comparison with the current process. A clear rule, better dashboard, or structured handoff may sometimes solve the problem more reliably than a complex model.
- Describe the decision, the person accountable for it, and the action that follows.
- Set a baseline using the current workflow before judging a new approach.
- Choose the simplest reliable method for the operational need.
Data quality is operational quality
Incomplete reservation histories, inconsistent property labels, missing maintenance outcomes, and unclear status changes limit what any model can learn. Bringing information into a usable structure often creates value before AI is introduced. It also makes later results easier to explain because the team knows what each field represents and where it originated.
Data access deserves the same care. A property workflow may contain personal or commercially sensitive information, so collection should be limited to the stated purpose and access should reflect role and need. Teams benefit from recording data owners, quality checks, updates, and exceptions rather than assuming an export is ready for automated use.
- Use consistent definitions for properties, reservations, statuses, and outcomes.
- Surface missing or conflicting information for review instead of silently filling gaps.
- Minimize personal data and apply role-based access.
Keep people accountable and manage risk continuously
A recommendation is not a guarantee. Pricing, tenant or guest communication, and other decisions affecting people need suitable review and controls. A practical system shows relevant inputs, records actions, supports override or escalation, and permits unexpected outcomes to be examined.
The voluntary NIST AI Risk Management Framework supports trustworthiness throughout the design, development, use, and evaluation of AI. Its Govern, Map, Measure, and Manage functions help teams establish accountability, understand context and risk, assess it, and respond over time. In property operations, that means documented purpose, ownership, monitoring, and correction when conditions change.
The aim is not to automate everything. It is to make the next operational decision better informed, easier to review, and more consistent without claiming a particular revenue or occupancy outcome.
- Provide override and escalation for consequential recommendations.
- Review performance and operational impact after release.
- Document assumptions, constraints, and non-use conditions.