When a checkout flow degrades at 2am, your on-call engineer needs three things in the next sixty seconds: which business transaction is affected, what the revenue impact looks like, and what worked the last time this happened.
Most observability platforms answer the first question. AI assistants take a swing at the third one, usually with confidence and often with hallucinations. Nobody connects all three.
That's the gap New Relic Knowledge closes.
RAG without business context is just search
Every observability vendor is shipping an AI assistant. Datadog has Bits AI. Dynatrace has Davis. Splunk has AI Assistant. They all promise grounded answers. Most of them ground in the same place: your documentation.
That's necessary, but it's not the differentiator.
New Relic Knowledge is built on something only New Relic has: a platform that already understands your business transactions, your service ownership, and the relationships across your entire stack. Every answer Knowledge surfaces is grounded twice. Once in your documentation. Once in the live business context that documentation is supposed to serve.
When an engineer asks "what should I do?", Knowledge doesn't just retrieve the runbook. It retrieves the runbook for the transaction that's currently degraded, with the historical incidents that touched the same service ownership boundary, weighted by what actually resolved the problem before.
That's the difference between AI that sounds correct and AI that is correct in your environment.
From "what broke" to "what it costs and what to do"
Observability platforms have been strong at detection for a decade. Mean time to detect is a solved problem for most teams. The hard problem is what comes next:
Is this a business-impacting event or a contained technical issue?
Which transaction is affected, and what's the revenue at stake?
What did we do the last three times this happened?
Which runbook applies to this service, not a similar one?
New Relic Knowledge answers all four, in the same view, in seconds. Surface similar past incidents weighted by entity relationship. Recommend remediation steps grounded in what worked. Retrieve the right runbook for the actual service, not a keyword match. Connect every answer back to the business transaction the engineer is trying to protect.
No tool switching. No context loss. No hallucinations dressed up as confidence.
Operational Resilience, measured in business outcomes
The "before" state most engineering orgs live in: incidents detected by customers before engineering, mean time to understand business impact measured in hours, alert noise overwhelming signal, no connection between service degradation and the transaction it supports.
The "after" state with New Relic Knowledge: when something breaks, the team knows within seconds whether it's business-impacting or contained. The right historical context surfaces automatically. Revenue impact is quantified before the war room convenes. The first action is informed by what actually worked, not by what sounds plausible.
This is what operational resilience means when it's anchored in business performance, not just system health.
Trust is a function of grounding, not tone
Engineers don't trust AI because it sounds confident. They trust it because they can verify the source.
Every Knowledge response cites the runbook, the past incident, the dashboard, the specific service entity it pulled context from. Engineers see the chain. They can check it. They can correct it. The system gets better.
This is also what makes Knowledge audit-ready as AI moves deeper into operations. When a regulator or a CTO asks why an action was recommended, the answer is a traceable chain back to your own operational history, not a black-box model output.
The foundation for autonomous operations
Knowledge is useful today. It's also the architectural foundation for what's coming next.
As engineering orgs move toward AI-driven automation, the quality of autonomous decisions depends entirely on the quality of the operational context those decisions are grounded in. A system that decides to roll back a deployment, throttle a service, or page a human needs to be grounded in your environment, your transactions, your history of what worked.
Knowledge is how you build that foundation. Every incident makes it more accurate. Every resolution makes it more reliable. Every team that uses it makes it a stronger asset for the agents that will increasingly operate alongside them.
Your operational history is already your competitive advantage. Make it accessible.
Your team has solved this incident before. Your runbooks have the answer. Your retrospectives captured the lesson. The work has been done.
New Relic Knowledge is how you make that work compound, in the moments when it matters most, grounded in the business context that makes it actionable.
Generally available May 25, 2026. At launch, Knowledge supports Confluence as a primary source, with Jira, GitHub, Slack, and additional integrations on the roadmap.
Ready to get started? Learn more here.
Don’t have a New Relic account yet? Sign up and get 100 GB + 1 user free. Forever. No credit card required.
Las opiniones expresadas en este blog son las del autor y no reflejan necesariamente las opiniones de New Relic. Todas las soluciones ofrecidas por el autor son específicas del entorno y no forman parte de las soluciones comerciales o el soporte ofrecido por New Relic. Únase a nosotros exclusivamente en Explorers Hub ( discus.newrelic.com ) para preguntas y asistencia relacionada con esta publicación de blog. Este blog puede contener enlaces a contenido de sitios de terceros. Al proporcionar dichos enlaces, New Relic no adopta, garantiza, aprueba ni respalda la información, las vistas o los productos disponibles en dichos sitios.
