Share with your CIO
A quarter of enterprise AI agents are running without any monitoring, according to New Relic’s 2026 Observability Forecast, which surveyed 2,575 IT and engineering leaders. That blind spot is serious: agents are autonomously modifying code, configurations, and infrastructure, which means unmonitored failures can compound quietly before anyone notices. The response is loud, with 80% of organizations planning to deploy AI-specific observability tools within 12 months, and vendors including AWS and Cisco already moving. Tool count has crept back up to an average of five per organization after years of consolidation effort.
What this means for your business
The 25% unmonitored figure is the number worth sitting with. If your organization has deployed agents across any meaningful set of workflows, the odds that at least one is operating without telemetry, the system data showing how a process is actually behaving, are not trivial. This isn’t a theoretical governance gap. Agents that rewrite configurations or alter infrastructure without observation trails create incident risk that’s harder to diagnose and slower to resolve than traditional software failures, because the failure state itself may be a consequence of a prior autonomous action no one logged.
New Relic has obvious commercial reasons to foreground the observability gap, and the report’s projection that 80% of enterprises will adopt AI observability tools tracks conveniently with the market it sells into. That tilt doesn’t make the underlying data wrong, but it does mean the tool-count framing deserves scrutiny. The drift from 4.4 tools in 2025 back to five in 2026 is presented as a warning, yet the diagnosis that unified platforms solve sprawl is exactly the pitch a full-stack observability vendor would make. The real consolidation pressure is real; the specific prescription benefits from independent validation.
High-impact outage costs dropped modestly, from $76 million to $73.5 million, but more than a third of organizations are still experiencing at least one major outage per week. That’s the number that reframes the observability budget conversation. If you’re heading into a renewal with a point-solution observability vendor, or defending the cost of a newer AI monitoring layer to the CFO, the $73.5 million average outage figure is the anchor, not the tool count. Organizations that added AI agent capabilities without extending their observability coverage first are already on the wrong side of that statistic.
Based on reporting from 1 in 4 agents run unmonitored, giving way to operational risk, originally published 2026-09-22 16:47:00.
