Work Services Skills Contact
Back to projects
AIAutomationProperty Management

Agente IA para Operaciones de Rentas Corta Estancia

End-to-end AI workflow automating guest message triage and response suggestions for short-term rental managers, reducing manual work by 70% at under $0.05 per execution.

The Challenge

Los property managers de rentas cortas estancia (Airbnb, VRBO) reciben decenas de mensajes diarios de huéspedes con preguntas repetitivas: check-in, WiFi, amenidades, problemas de mantenimiento. El triage manual consumía más del 70% del tiempo operativo.

Cada mensaje requería lectura, clasificación por urgencia, y una respuesta personalizada. En temporada alta, esto escalaba a cientos de mensajes diarios, imposibles de manejar sin contratar personal adicional.

Rental AI Flow - conceptual diagram showing message processing pipeline
Conceptual flow: User message → Clean → AI processing → Format verification → Excel logging

What We Built

Diseñamos un workflow end-to-end en n8n con integración de IA generativa:

n8n workflow showing complete AI agent architecture with guardrails and Google Sheets integration
Complete n8n workflow: Webhook → Guardrails → LLM Chain → OpenRouter → Output validation → Google Sheets
  • Webhook trigger que captura mensajes de huéspedes en tiempo real
  • Guardrails de seguridad para filtrar contenido inapropiado antes de procesar
  • Clasificación y priorización con IA (OpenRouter) por categoría: urgente, consulta, mantenimiento
  • Generación de respuestas sugeridas contextualizadas al huésped y la propiedad
  • Human-in-the-loop: las respuestas críticas requieren aprobación humana antes de enviarse
  • Audit trail completo en Excel con cada interacción, decisión y costo
  • Results

    Reducción del 70% en tiempo de triage manual. Los property managers ahora dedican su tiempo a tareas de alto valor en lugar de responder preguntas repetitivas.

    Costo de ejecución menor a $0.05 por corrida (promedio de 5 llamadas de IA). El sistema es económicamente viable desde el primer día.

    Trazabilidad completa: cada mensaje, clasificación y respuesta queda registrada en Excel para auditoría y mejora continua.

    Asistente Rental Flow UI showing structured AI response with category, priority, and JSON output
    Live demo interface: AI classifies messages by category and priority, suggests responses, and logs structured JSON data

    See It In Action

    Tools & Technologies

    • n8n Workflow Automation
    • OpenRouter (Multi-model AI)
    • Webhooks & REST APIs
    • Excel (Audit Trail)
    • Human-in-the-loop Architecture

    Architecture Decisions

    Why OpenRouter instead of direct API calls? OpenRouter provides access to multiple LLM providers (OpenAI, Anthropic, Mistral) through a single API. This allows me to switch models based on cost-performance ratios without changing the workflow structure. For this use case, I use GPT-4 for complex classification tasks and GPT-3.5 for simple response generation, optimizing cost while maintaining quality.

    Why Excel instead of a database? Property managers already live in Excel. They know how to filter, sort, and analyze spreadsheets. By using Excel as the audit trail and configuration layer, I eliminated the learning curve. The property manager can see every decision the AI made, override classifications, and adjust response templates without touching code.

    Why human-in-the-loop for critical actions? Fully autonomous systems work until they don't. For maintenance requests involving expensive repairs or guest complaints that could lead to bad reviews, I built a approval workflow. The AI drafts the response, but a human reviews and approves before sending. This catches edge cases and builds trust with the property manager over time.

    Cost Analysis

    Operating cost per message: $0.03 to $0.08 depending on message complexity and model selection. A typical property manager handling 100 messages per day spends $3 to $8 daily on AI processing. Compare this to hiring a virtual assistant at $15/hour who can handle 30 messages per hour ($0.50 per message).

    Monthly infrastructure cost: The n8n workflow runs on a $20/month VPS with 2GB RAM and 2 vCPUs. This handles up to 10,000 messages per month. For larger operations, I scale horizontally by adding more workers, keeping the cost per message constant.

    ROI calculation: A property manager spending 30 hours per week on message triage at $25/hour = $750/week in labor costs. The automated system costs $50/week to operate. That's a 93% reduction in operational costs, or $2,800/month in savings.