From AIOps to Autonomous Operations: An Agentic AI Architecture for Self-Healing Enterprise IT Infrastructure
Year: 2026
Type: article
Abstract: Enterprise IT estates spanning data centres, multiple clouds, networks and end-user devices have outgrown reactive operating models. AIOps platforms have improved event correlation and anomaly detection, yet in most enterprises they stop at insight, leaving diagnosis and remediation to engineers. This paper proposes a vendor-neutral reference architecture that extends AIOps towards autonomous, self-healing operations using agentic AI. The architecture unifies observability data, IT service management records, configuration and topology information and historical incident knowledge into an operational data layer. Specialised agents for detection, correlation, root cause analysis, remediation planning, execution and validation work on this layer. Large language models support reasoning and runbook interpretation, while every action runs through governed automation workflows. A four-stage maturity model is defined, progressing from reactive to predictive, preventive and autonomous operations, along with governance controls such as risk-based approval thresholds, human-in-the-loop checkpoints, change management integration, rollback and audit trails. Application scenarios across Service Desk, End User Computing, Data Centre, Network and NOC operations are discussed. The approach offers enterprises and managed service providers a practical path to reduce manual effort, shorten incident resolution and move IT operations from firefighting to proactive service assurance.
Author: Mihir Patel
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7225911270
