Tenant Service Request Triage and Resolution Automation

This agent acts as the unified front door, ingesting requests from chat, email, tenant portals, and voice (via transcription). It uses fine-tuned NLP models to classify the request type (e.g., HVAC, plumbing, access), urgency, and required trade. By extracting entities like unit number and asset ID, it structures unstructured data for downstream systems, eliminating manual data entry and misrouting.

Before routing, this component enriches the request by querying integrated systems. It retrieves the tenant's lease terms, service-level agreements (SLAs), and the asset's maintenance history from the CMMS (e.g., ServiceNow). It also searches a vector-based knowledge base for known solutions (e.g., 'AC filter reset instructions'). This context allows the system to auto-resolve common issues or provide precise information to the assigned technician.

The core decision logic uses classified intent, retrieved context, and real-time data (technician GPS, parts inventory) to determine the action. It can: 1) Auto-Resolve with instructional messages, 2) Create & Dispatch a prioritized work order to the CMMS, assigning the optimal internal team or vendor based on skill, location, and SLA, or 3) Escalate complex or high-risk issues to a human agent with a full case summary. Integration with vendor portals enables two-way status sync.

To close the loop with tenant experience, this agent updates the property's CRM (e.g., Salesforce) with the full interaction history and work order status. It also triggers proactive, multi-channel notifications (SMS, app, email) to the tenant—confirming receipt, providing ETA updates, and soliciting post-resolution feedback. This automation reduces inbound status inquiry calls by over 80% and builds tenant satisfaction through transparency.

Critical for governance and handling exceptions. The workflow includes configurable approval rules for actions exceeding cost thresholds, requiring lease modifications, or involving access to sensitive areas. Non-standard requests or low-confidence AI classifications are routed to a human review queue within the property management platform, with all context pre-assembled. This ensures control is maintained while the AI handles the routine.

The implementation is not static. All decisions, response times, and resolution outcomes are logged to an observability platform. Dashboards track KPIs like first-time fix rate, SLA compliance, and auto-resolution success. This data feeds a continuous training loop where misclassifications are used to retrain the NLP models, and routing inefficiencies are used to refine the dispatch logic, creating a system that improves operationally over time.

The primary business owner who defines the strategic objectives: reducing call center volume by 40%, improving tenant satisfaction (NPS/CSAT), and lowering operational cost per request. This role approves the budget, champions the change with leasing and facilities teams, and defines the key performance indicators (KPIs) for success, such as mean time to resolution (MTTR) and auto-resolution rate.

Owns the end-to-end system architecture. Designs the multi-agent orchestration (e.g., using LangGraph or CrewAI), specifies integrations with the CRM (Salesforce), tenant portal, and building management systems, and defines the data flows for NLP classification, knowledge base retrieval, and work order creation. This role ensures the system is scalable, secure, and maintainable, selecting appropriate cloud services, APIs, and observability tools.

Responsible for developing, training, and maintaining the core intelligence layer. This includes fine-tuning the NLP model for request classification (e.g., maintenance, billing, access), building the RAG pipeline for lease and FAQ retrieval, and implementing the logic for the resolution agent (e.g., providing self-help instructions, scheduling a ticket). They establish model monitoring for accuracy drift and manage the CI/CD pipeline for agent updates.

The primary operational stakeholder for the maintenance workflow. Defines the service level agreements (SLAs), ticket prioritization rules (emergency vs. routine), and the dispatch logic for internal crews vs. external vendors. This role validates the automated work orders generated by the system and is responsible for the human-in-the-loop review queue for complex or safety-critical requests that cannot be auto-resolved.

Manages the deployment, infrastructure, and governance. Provisions the cloud environment, containerizes agents, establishes the CI/CD pipeline, and ensures robust logging and monitoring (e.g., Datadog, Grafana). This team implements security controls for data privacy (tenant PII), manages API keys, and ensures the system complies with enterprise IT policies and regulatory requirements (e.g., data residency).

Owns the user adoption and communication strategy. Develops training materials for property staff on using the new dashboard and handling escalations. Designs and executes the tenant rollout plan, including communications about the new chat/portal interface and managing the transition from phone-heavy support. This role collects feedback, measures tenant satisfaction, and iterates on the service design based on operational data.