AI observability gap: a quarter of enterprise AI unmonitored in production

What happened

New Relic's 2026 Observability Forecast estimates that roughly a quarter of enterprise AI workloads are running in production without monitoring, according to coverage by Diginomica. The report frames this unmonitored share as a measurable gap rather than a theoretical risk.

The findings arrive alongside a leadership change at the company. Diginomica reports that Nandini Ramani is now New Relic's chief executive, taking over a business whose observability products are already positioned to address this kind of visibility gap.

Why it matters

As organisations move AI models and automated agents into live operations, the question shifts from whether they work in testing to whether anyone can see how they behave in production. An unmonitored system can degrade for weeks before a customer, employee or operator notices, which means the first signal of failure is often a poor experience rather than an alert.

For leaders in experience, AI and digital transformation, this points to observability as an operational prerequisite for AI adoption, not an afterthought. It also suggests that governance of AI services will increasingly depend on monitoring capability that sits alongside, rather than behind, deployment decisions.

By the numbers

• About a quarter of enterprise AI running in production is reported to be unmonitored, according to New Relic's 2026 Observability Forecast as covered by Diginomica.
• 2026 is the forecast year for the report's findings.

The Renascence take

Most readers will treat this as an engineering statistic. We think it is a service-design problem. When a quarter of AI is invisible, the people who pay for that invisibility are customers and frontline staff, who absorb errors that nobody has been measuring.

Unmonitored AI is not a neutral gap; it is an experience risk with no feedback loop. The behavioural principle here is simple: you cannot improve what you cannot see, and customers are usually the first sensors. Operators should treat observability as part of the service contract, with defined experience metrics for every AI-driven touchpoint, before they expand deployment. A new CEO inheriting a product built for this gap is a reminder that the market has already decided visibility matters.