New Relic’s latest observability report finds that roughly a quarter of enterprise AI workloads are running in production without adequate monitoring. Data platform leaders across the DACH region tell us the same thing: the pressure to ship AI features has outpaced the tooling and expertise to observe them once live.
The report lands alongside a leadership change. Nandini Ramani takes over as CEO, inheriting a product portfolio positioned around this monitoring gap. Companies building AI into core operations need people who can tell them what their models are actually doing in production.
Observability expertise now appears as a core requirement in platform engineer specifications we receive. Clients want candidates who sit between data science teams and infrastructure, someone who understands both the ML pipeline and the production environment well enough to know what to instrument and why.
Traditional observability grew up around web applications and microservices. Applying it to AI workloads, where outputs are probabilistic and failure modes are harder to define, requires a different mental model. Candidates who have actually done this work stand out immediately in screening conversations compared to those who have only read about it.
A data platform lead at a Munich manufacturer told us last month that finding someone who could instrument their recommendation engine properly took twice as long as hiring the ML engineers who built it. The expertise exists, but it sits with people who learned it inside hyperscalers or observability vendors. Reaching them requires targeting passive candidates who are excelling in those environments.
Hiring managers building data platform teams should treat observability as a first-class requirement when adding AI capability. The candidates who can bridge inference pipelines and production monitoring command a premium, and the search requires looking beyond the usual applicant pools.
Prompted by reporting from Diginomica.