How does the world’s largest peer-to-peer car sharing marketplace manage a high-velocity engineering organization running cloud infrastructure, microservices, and Kubernetes clusters at scale?
For Turo, that means moving beyond local laptop coding and deploying context-aware AI agents directly into production workflows.
Turo’s on-call engineers faced high toil and increased MTTR when triaging alerts for microservices they were unfamiliar with. At the same time, high-volume AI code generation created downstream bottlenecks in deployment pipelines and code reviews, shifting the engineering focus from simply writing code to accurately defining the problem.
When generative AI tools emerged, Turo raced ahead, equipping every engineer with local AI coding assistants. The ability to write code faster was only just beginning, but that meant new problems popped up: downstream bottlenecks surfaced in code review and release pipelines soon after.
New Relic’s VP of Developer Relations Jemiah Sius spoke with Turo’s VP of Engineering Victor Mora for a fireside chat to share how Turo uses New Relic to power cloud AI agents, slash on-call triage times, and safely scale engineering velocity.
Here’s what they uncovered.
Key Takeaways
- Move Beyond IDEs: How Turo evolved from local coding assistants to running cloud AI agents directly in production.
- 60-Second Incident Triage: How New Relic MCP server lets Turo’s AI agents diagnose critical outages before on-call engineers even log on.
- The Shift to Spec-Driven Dev: Writing code is no longer a bottleneck but code review and deployment are, and how automated canary testing keeps release pipelines moving quickly.
Why Fast Code Creates Slow Pipelines
Turo’s operational shift mirrors a massive trend across the software industry. In New Relic’s 2026 State of AI Coding Report, 67% of technology leaders revealed that AI now generates or significantly refactors between 51% and 75% of their organization's weekly code output.
However, the report uncovered a major paradox:
- 94% of leaders rate AI-generated code as higher quality during initial review
- 78% report a surge in production incidents after shipping AI-generated code
- 86% experienced increased senior-engineer firefighting
Why does this happen? AI coding tools have full visibility into source code syntax but zero visibility into real-world production traces, dependencies, and application state.
Writing code is the easy part. Now the hardest part is: what is the problem that you're trying to solve with code? We write so much code, but code review and deployment pipelines—especially for our monolith—are becoming the productivity bottleneck."
At the same time, when production incidents hit, engineers faced significant fatigue trying to triage alerts across microservices they didn't routinely work on. Turo needed a way to give AI tools full production awareness so they could assist engineers where it mattered most: incident response, code safety, and system design.
Giving Cloud AI Agents Direct Sight into Production
AI agents are only as good as the context they are fed. Without live telemetry, an AI agent operates blindly. New Relic provided Turo with the ideal foundation to fix this.
- A Unified Telemetry Hub: Turo streams all cloud infrastructure, Kubernetes metrics, application logs, and deployment events into New Relic, creating a single source of truth.
- The New Relic MCP (Model Context Protocol) Server: By connecting cloud-hosted AI agents directly to New Relic via MCP, agents can query live production telemetry using natural language without requiring engineers to manually write complex NRQL (New Relic Query Language) queries.
- Continuous Deployment Safety: New Relic tracks feature flag toggles and deployment events in real time, giving AI agents immediate context on whether a recent code change caused a system anomaly.
60-Second Root Cause
The full power of this integration was discovered during a recent critical midnight incident at Turo.
Under traditional on-call rotations, a midnight alert forces an engineer out of bed and into action to discover which service has failed as quickly as possible.
With New Relic and cloud-based AI agents, the workflow transformed completely. Here’s what happened within the first 60 seconds of the alert firing:
- The AI agent queried New Relic
- Analyzed the service connections and log paths
- Identified the exact root cause
- And generated an action-ready diagnostic report
The best part? All of this happened before the engineer even had time to log on. The team mitigated the issue immediately, dramatically reducing MTTR and protecting Turo’s customer experience.
Grounding Autonomous AI in Production-Grade Observability
Turo's success reinforces a modern requirement for engineering teams: you cannot control AI agents without complete observability.
As AI code generation accelerates, closing the gap between pull request reviews and production reality is one of the most important decisions engineering leaders will make this year.
Transform raw telemetry into trusted operational intelligence with Ground Truth, New Relic’s operational context layer. Through native MCP tools, Ground Truth equips your AI agents with full-stack visibility, turning complex observability data into immediate, accurate action.
And for all the latest in observability trends, download the 2026 State of AI Coding Report to explore insights from 200 technology leaders on managing agent debt, code quality, and production risk.
本ブログに掲載されている見解は著者に所属するものであり、必ずしも New Relic 株式会社の公式見解であるわけではありません。また、本ブログには、外部サイトにアクセスするリンクが含まれる場合があります。それらリンク先の内容について、New Relic がいかなる保証も提供することはありません。
