Tech Transformed

How Do You Monitor AI Agents in Production Without Breaking Incident Response?

February 18, 2026·22 min·1 clip
"If it's not observed, then it's rogue"—Manish Taylor ties AI governance to observability and compliance.
1. Tech Transformed asks how to monitor AI agents in production, as Shubhangi Dua and Manish Taylor discuss AI-driven development, outages, and full-stack observability. 2. Dua is the host and EM360 Tech producer, while Taylor is New Relic's EMEA Field Chief Technology Officer, and both frame the episode around operational resilience and developer efficiency. 3. Taylor's core warning is simple: if you do not get your "observability house in order," then AI innovation, production reliability, and compliance are at risk. 4. He says organizations are building solutions faster through AI-generated code and "wipe coding," but many prototypes and POCs now reach production before due diligence. 5. That speed, Taylor argues, raises cognitive load on operators, because systems were not designed for scaling, human factors, or long-term operation. 6. He contrasts monitoring for "known knowns" with observability for "unknown unknowns," saying microservices, SaaS components, and AI-built tools create more fragmentation and more noise. 7. Taylor adds that data observability matters too, because structured and unstructured data both need quality checks, accuracy, and mining inside AI systems. 8. He says automation and retrieval augmented generation can gather more information from systems of knowledge and records, arming experts to make better decisions faster. 9. Taylor's New Relic example shows the company using a production instance to monitor itself, and an internal SRE agent to predict issues before customer-facing impact. 10. CTOs, CIOs, and observability teams would get the most from this episode.

As heard by us

A practical PlayNext discussion on why AI systems need observability before production failures become harder to contain.

AI is changing the ways enterprise software can break, and this PlayNext episode treats observability as the discipline that keeps that change from outrunning operational resilience.

Read the full review in PlayNext →

Why you'd press play

Press play if your incident process still assumes AI-driven software will fail in familiar ways.

Read the full recommendation in PlayNext →
Listen to the show on