Orion AI: How Cornerstone OnDemand Cut Database Diagnosis by 78% with a Three-Agent System on Amazon Bedrock

Описание и применение

Cornerstone OnDemand внедрил Orion AI, систему из трёх специализированных агентов, работающих на Amazon Bedrock, чтобы сократить время диагностики баз данных с 45 до 10 минут, автоматизировав 60 % инцидентов и обрабатывая более 40 оповещений в день.

Задача

Manual database diagnosis at scale took an average of 45 minutes per alert, consuming most engineers’ time on repetitive tasks.

Результат

In six months, Orion AI reduced diagnosis time by 78 %, automated 60 % of incidents, and enabled handling of 40+ alerts daily with structured audit trails and approval gates.

Рабочий процесс

• 1. Alert ingestion via EventBridge
• 2. Diagnosis Agent analyzes alert data and metrics
• 3. Query Generation Agent creates safe, read‑only SQL and validates syntax
• 4. Executed queries via Lambda with safety checks
• 5. Remediation Agent proposes actions and requires human approval for destructive ops
• 6. Approved fixes executed within transaction boundaries
• 7. All steps logged for audit and observability

Доказательства

Cornerstone OnDemand built Orion AI to turn database operations from reactive firefighting into proactive automation.

Источник

A three-person team cut database diagnosis from 45 minutes to 10 minutes in six months using Amazon Bedrock and the Strands Agents framework.

Источник

Diagnosis time dropped from 45 minutes to 10 minutes (78% reduction)

Источник