→ Back to Home
Observability

Gartner Warns 40% of Organizations Scaling AI Face Business-Critical Disruptions by 2028 Due to Weak Governance

Gartner announced today at its IT Infrastructure, Operations and Cloud Strategies Conference in London that a staggering 40% of infrastructure and operations (I&O) organizations scaling AI will suffer a business-critical service disruption by 2028 if they fail to implement strong runtime governance. This projection is based on a survey of 298 I&O leaders in the US and UK, conducted between June and July 2026, which found that 95% of organizations have already experienced at least one incident related to AI agents in the past year. These incidents include cascading API loops, unauthorized configuration drift, and cloud resource exhaustion. This finding is critical for any practitioner involved in deploying or managing AI systems, especially those leveraging autonomous agents. The immediate implication is that the perceived safety of AI agent operations (70% of I&O leaders believe their agents operate safely) is dangerously misaligned with reality. The report highlights a severe operational readiness gap, indicating that current governance frameworks are weak (57%) and existing IT risk tools are outdated (55%) for handling autonomous machine behaviors. Without proper guardrails, the promise of AI-driven efficiency could quickly turn into widespread instability and costly outages. This trend fits squarely within the broader industry push towards autonomous operations and AI-driven automation. While AI copilots have been prevalent for several years, the current wave sees a rapid acceleration in the deployment of agentic architectures with delegated execution authority. These agents directly interact with APIs, databases, and command shells, making robust governance and observability paramount. The industry has been moving towards more intelligent, self-managing systems, but this report serves as a stark reminder that the control planes and safety mechanisms have not kept pace with the rate of AI adoption. Other recent reports, such as Dynatrace's State of Observability 2026, also highlight the gap between AI ambition and execution, noting that only 27% of organizations address visibility into how agents decide and act. In practice, practitioners must immediately inventory all active AI agents and classify them as non-human identities within their identity platforms. This requires a fundamental shift from traditional automation governance to a proactive, agent-specific infrastructure governance approach. Key actions include establishing policy controls, permission models, enhanced observability instrumentation to track agent reasoning chains, and audit capabilities. Crucially, only 15% of organizations currently have automated circuit breakers to halt runaway agent execution, a figure that needs to drastically increase. Teams should focus on implementing real-time guardrails that can slow down misbehaving agents before they cause disruptions and utilize observability to understand *why* an agent took a particular action. Ignoring these warnings will likely lead to significant business disruptions and erode trust in AI initiatives.
#ai governance#observability#autonomous agents#devops#it operations#risk management
Read original source