ID:83500さん

Project Overview: ServiceNow EC Portal Optimization via Agentic AI Role: Technical Project Manager Project Scope: Implementation of an autonomous Agentic AI framework integrated into the ServiceNow Employee Center (EC) portal to automate complex IT and HR service request fulfillment, routing, and real-time troubleshooting. The Problem: The legacy ServiceNow portal suffered from significant response downtime and high ticket resolution latency, primarily due to rigid, rule-based workflows and manual bottlenecks. The Goal: Deploy autonomous AI agents capable of understanding user intent, executing end-to-end multi-step fulfillment processes, and reducing response downtime. Key Achievements & Impact Reduced Response Downtime: Cut overall portal response and resolution downtime by 42% within the first quarter post-launch. Automated First-Contact Resolution: Achieved a 35% deflection rate of Tier-1 support tickets via autonomous agent intervention. Enhanced Service Continuity: Maintained a 99.9% availability rate for the AI middleware, ensuring 24/7 self-service capabilities for global employees. Strategic Challenges & Leadership Solutions 1. Cross-Cultural Leadership & Communication Friction The Challenge: The project team was highly distributed, comprising a core AI engineering squad in Bangalore, ServiceNow architects in Europe, and business stakeholders in the US. Misalignments in communication styles and contrasting working cultures initially led to missed dependencies in the API integration phase. How I Overcame It: I established a "Unified Shared Context" framework. I synchronized core sprint ceremonies to overlap cross-cultural time zones and replaced text-heavy requirement docs with visual, interactive architectural maps. By fostering a psychologically safe environment where team members were encouraged to clarify assumptions explicitly, I bridged the communication gap and unified the team under a singular technical roadmap. 2. Guardrailing "Agent Drift" and Non-Deterministic AI Behaviors The Challenge: Unlike traditional chatbot scripts, Agentic AI operates semi-autonomously, utilizing LLMs to dynamically orchestrate its own tool calls. During staging, we encountered "agent drift"—where the AI generated infinite loops or unexpected API calls inside ServiceNow when faced with ambiguous user prompts. How I Overcame It: I shifted the team toward a "Human-in-the-Loop" (HITL) and deterministic guardrail strategy. I managed the implementation of strict semantic validation layers and confidence scoring thresholds. If an agent's planned action sequence scored below an 85% confidence interval, the system seamlessly and invisibly rolled back the transaction and escalated the context to a human agent, guaranteeing portal stability. 3. Legacy Architecture Integration Bottlenecks The Challenge: The existing ServiceNow EC Portal configuration featured heavily customized legacy workflows that did not natively support streaming APIs or asynchronous webhooks required by the Agentic AI orchestration layer. How I Overcame It: I led a technical deep-dive with the ServiceNow engineering leads to design a lightweight, decoupled middleware layer. This architecture acted as a high-speed translation bridge between the non-deterministic AI agents and ServiceNow’s rigid REST APIs, preventing portal performance degradation. PM Takeaway: By balancing rigorous technical guardrails with adaptive cross-cultural leadership, we successfully transformed a legacy service catalog into an intelligent, self-healing ecosystem—significantly boosting employee satisfaction scores across the enterprise.