ActionNex: A Virtual Outage Manager for Cloud Computing
Organizations: Microsoft PRIMO, Redmond, USA · Microsoft Azure Core, Redmond, USA · Microsoft Azure CTO Office, Redmond, USA · Microsoft Research, Redmond, USA
Abstract
Outage management in large-scale cloud operations remains heavily manual, requiring rapid triage, cross-team coordination, and experience-driven decisions under partial observability. We present ActionNex, a production-grade agentic system that supports end-to-end outage assistance, including real-time updates, knowledge distillation, and role- and stage-conditioned next-best action recommendations. ActionNex ingests multimodal operational signals (e.g., outage content, telemetry, and human communications) and compresses them into critical events that represent meaningful state transitions. It couples this perception layer with a hierarchical memory subsystem: long-term Key-Condition-Action (KCA) knowledge distilled from playbooks and historical executions, episodic memory of prior outages, and working memory of the live context. A reasoning agent aligns current critical events to preconditions, retrieves relevant memories, and generates actionable recommendations; executed human actions serve as an implicit feedback signal to enable continual self-evolution in a human-agent hybrid system. We evaluate ActionNex on eight real Azure outages (8M tokens, 4,000 critical events) using two complementary ground-truth action sets, achieving 71.4% precision and 52.8-54.8% recall. The system has been piloted in production and has received positive early feedback.
Figures & tables
| Memory | Source & Content | Role in Reasoning |
|---|---|---|
| Long-Term Memory (Knowledge Base) | Aggregated reference knowledge, including outage management manuals, SOPs, service playbooks , guidelines, and domain-specific expertise, indexed as Key–Condition–Action (KCA) triples. | Provides canonical solutions and best-practice actions for known scenarios when matching conditions are satisfied. |
| Episodic Memory (Case Library) | Logs of historical outages , including critical events, decisions taken, and outcomes (success or failure). Each case records the outage context (e.g., service, severity), the identified precondition, and the actions used for resolution. | Provides analogous examples from prior outages. Successful actions suggest viable strategies, while failed actions indicate approaches to avoid. |
| Working Memory (Current Context) | Live Outage state , including recently attempted actions, intermediate results, conversation history, and the current outage stage or status. This memory is continuously updated as the outage evolves. | Provides situational awareness , ensuring recommendations account for what has already occurred, thereby preventing redundant or contradictory actions and maintaining a coherent action sequence. |
| Ground Truth | Precision (%) | Recall (%) |
|---|---|---|
| G1 | 71.4 | 52.8 |
| G2 | 71.4 | 54.8 |
AI-Assisted Decision Making
AI-assisted decision making studies how computational systems can inform human choices while accounting for context, uncertainty, and consequences.
See all 144 papers