Agentic Triage for Managing Planned Outage Notifications and Customer Reschedules

Manual processes for identifying affected customers, drafting notifications, and managing reschedule requests consume hundreds of agent-hours per outage. A custom agentic workflow automates the entire sequence: pulling work plans from the OMS, cross-referencing addresses in the CIS, generating personalized notifications, and handling inbound rescheduling via chat or IVR. This cuts notification cycle time by over 70% and proactively manages customer expectations, reducing complaint volumes by 40-60%.

Missing mandated advance notice windows for planned outages results in regulatory fines and customer credits. An autonomous workflow embeds compliance logic, calculating legal notice deadlines based on outage type and jurisdiction. Agents automatically trigger notifications the moment the work plan is approved, with audit trails proving delivery. This transforms compliance from a manual checklist into a system-enforced control, eliminating penalty risk and ensuring consistent policy application across thousands of customers.

Unmanaged customer reschedule requests lead to crew downtime, inefficient routing, and project delays. The workflow includes an intelligent rescheduling agent that evaluates crew capacity, travel time, and job dependencies in the workforce management system. It negotiates alternative windows with customers via automated dialogue, only confirming slots that maintain overall schedule integrity. This turns rescheduling from a disruptive cost center into a controlled optimization lever, protecting crew utilization and project timelines.

For outages affecting medical facilities, industrial processes, or vulnerable populations, generic notifications are insufficient and risky. The workflow agents segment customers using CIS flags and historical data, then orchestrate tiered, multi-channel communications (SMS, IVR, direct call) with escalation paths. For critical cases, the system can initiate proactive outbound calls or escalate to a human agent with full context. This builds customer trust and operational resilience by demonstrating control over high-stakes scenarios.

Value is lost if the automation creates new data silos. The implementation architecture focuses on a lightweight orchestration layer (e.g., LangGraph) that brokers real-time data between the OMS, CIS, MDMS, and field scheduling systems. For legacy systems without APIs, browser-automation agents simulate user inputs. This approach delivers a unified operational view without a costly core system replacement, turning fragmented data into a single source of truth for outage communications.

Autonomous customer communication requires robust controls. The workflow design includes approval gates for initial notification batches, real-time monitoring dashboards for agent performance, and immutable logs of all customer interactions and system decisions. Exception handling routes complex cases to human agents with full context. This governance model ensures the operation is controlled, explainable to regulators, and can be continuously tuned based on performance data, moving from an ad-hoc process to a managed service.

This agent pulls structured work plans from the Outage Management System (OMS) or Enterprise Asset Management (EAM) system via API. It parses the affected circuits, transformers, or addresses, then cross-references them against the Customer Information System (CIS) to build a precise list of impacted accounts. It filters for critical care customers and those with sensitive processes (e.g., medical equipment, manufacturing) for prioritized handling.

An orchestration agent manages the delivery of advanced outage notices across SMS, email, IVR, and customer portal alerts. It personalizes messages with outage windows, reason codes, and regulatory-mandated notice periods. The agent handles delivery status, retries failed attempts, and logs all communications for audit trails. It integrates with telecom APIs and email service providers for scalable, reliable dispatch.

A conversational agent fields inbound customer requests to reschedule service interruptions. It accesses the workforce management system to check crew availability and feasibility. Based on business rules (e.g., cannot delay critical infrastructure work), it either confirms a new window, offers a limited alternative, or escalates to a human scheduler. It updates the OMS and CIS in real-time to maintain a single source of truth.

This control layer ensures the workflow adheres to public utility commission regulations for notice periods and critical customer protocols. It injects mandatory hold periods between notification and outage start, routes exceptions for legal review, and generates a immutable log of all agent decisions, customer interactions, and system actions. This creates a defensible audit trail for regulatory reporting and operational review.

Not all requests can be handled autonomously. This component manages a prioritized queue for cases requiring human intervention: complex rescheduling conflicts, customer disputes, or system integration failures. It enriches each ticket with full context from previous agent steps and routes it to the appropriate team (e.g., customer care, field supervision) via ServiceNow or a similar ITSM platform, ensuring no case is dropped.

The workflow is built as a directed graph using LangGraph, where each node is a specialized agent or a conditional branch. It integrates via REST APIs and message queues with the OMS (Oracle NMS, Schneider Electric), CIS (SAP, Oracle CC&B), and workforce management systems. A central orchestration layer manages state, handles errors, and ensures data consistency across these legacy and modern systems, deployed in a containerized environment for scalability.