Enterprise AI Support Assistant — Multi-Agent Root Cause Analysis

Overview

An enterprise-grade multi-agent AI support assistant that automates production support investigations for large-scale data platforms. Instead of manually checking databases, Kubernetes clusters and application logs, engineers submit an issue through Microsoft Teams and the system performs a structured investigation, identifies the most likely root cause, assesses operational risk and generates step-by-step remediation guidance.

The problem

Production support engineers spend significant time switching between databases, Kubernetes clusters, monitoring tools and application dashboards before identifying the root cause of an issue. The process is repetitive, time-consuming and dependent on senior engineering expertise.

How it works

A support request arrives through Teams and is routed through specialised agents, each with a dedicated responsibility:

• Investigation Coordinator Agent — plans and directs the investigation.
• Database Investigation Agent — read-only SQL investigation.
• Kubernetes (AKS) Investigation Agent — cluster-level evidence gathering.
• ICEDQ Investigation Agent — platform-specific checks.
• Evidence Analysis & Root Cause Agent — correlates the collected evidence.
• Risk Assessment Agent — evaluates operational risk.
• Solution Recommendation Agent — produces step-by-step remediation guidance, reported back into Teams.

Safety and key engineering decisions

The security model is the defining constraint. Every agent operates with read-only permissions, hook-based authorization gates what may run and a fail-closed validation model ensures the AI cannot modify production systems. Secrets are redacted automatically from collected evidence. The AI recommends solutions; all operational decisions remain under human control.