AI System for Preparing Hotel Revenue Scenarios for Review

The problem

When occupancy, booking pace, event and channel data are held in different sources, staff must match fields manually before preparing a revenue scenario with explicit assumptions.

The revenue manager must compare every suggestion in the revenue scenario with explicit assumptions against the source record; incomplete or conflicting evidence must not replace the final decision.

If an unconfirmed group allocation is not set aside for separate review, the scenario could be treated as a definitive forecast; the record must not proceed without a human decision.

What we set out to improve

• Time taken for the draft revenue scenario with explicit assumptions, together with its supporting evidence, to reach the revenue manager for review
• Outcomes of records sent for human review because a group allocation is not confirmed
• Number of suggestions corrected after identifying a risk that the scenario could be treated as a definitive forecast

How the system works

1. 01 Retrieve occupancy, booking pace, event and channel data from permitted sources
2. 02 Verify field and document versions
3. 03 Prepare the revenue scenario with explicit assumptions and supporting evidence
4. 04 Check the exception: a group allocation is not confirmed
5. 05 Have the revenue manager accept, correct or reject the suggestion
6. 06 Decision gate: room rate, inventory and sales restrictions
7. 07 Record the decision and its source separately

Methods we used

• A source-identified data schema for occupancy, booking pace, event and channel data
• A structure that separates deterministic rules from AI suggestions
• A screen showing the revenue scenario with explicit assumptions alongside the supporting evidence
• A human review queue for unconfirmed group allocations
• A record identifier that prevents duplicate processing

Where people stay involved

Data and security

Only the necessary fields from occupancy, booking pace, event and channel data should be processed. Role-based access should be applied, and unnecessary identity, health, financial and internal organisational data should not be sent to the model. Source access, the AI suggestion and the revenue manager's decision should be recorded separately.

Who this suits

Frequently asked questions