77% of Enterprises Have Paid a Price for Ungoverned AI Agents

A Redwood Software survey of 300 technology, risk and AI leaders found 77% of enterprises had experienced a negative consequence from an ungoverned AI agent, most commonly unexpected cloud costs, followed by operational disruption and compliance issues. Agentic AI deployment grew at 89% of respondents’ organizations over the past year, but only 42% reported strong integration with legacy or core systems, and 42% had a centralized registry of deployed agents. The survey points to integration, visibility and fragmented ownership as governance challenges; 62% said better observability would help them scale agentic AI with confidence.

Analysis

The investable signal is less “AI agents are unsafe” than that production deployment is shifting the bottleneck from model access to integration, observability and control. That favors vendors that can govern workflows across heterogeneous systems—enterprise automation/orchestration, IT operations monitoring, identity and security—while raising the risk that AI budgets are diverted from experimentation into integration and controls. Potential beneficiaries include ServiceNow, IBM, UiPath, Datadog and Dynatrace, but hyperscalers can bundle overlapping capabilities, limiting standalone pricing power. Cloud providers may see near-term usage lift from poorly controlled agents, yet cost surprises could trigger tighter customer budgets and optimization later; that is a less durable benefit than the headline usage trend suggests.

Treat the evidence as directional, not a demand forecast: it is a vendor-sponsored survey of 300 respondents, and Redwood’s customer comparison does not establish that its product caused better preparedness. Near term, the release alone is unlikely to justify a sector position. Over 1–3 months, watch enterprise software and cloud earnings for commentary on agent-related production deployments, integration spending and cloud cost controls. Over 6–18 months, the structural winners are likely to be vendors with cross-platform governance and measurable incident prevention, unless Microsoft, AWS or Google make these controls native and commoditize independent tools. The contrarian risk is that buyers standardize on existing platforms rather than fund a new orchestration layer.

Key Decisions for Investors

• No trade on this survey alone. Use earnings calls and customer disclosures to verify whether governance/orchestration is producing incremental budgets rather than being absorbed into existing IT spend.
• Watch for a relative-value opportunity: favor diversified observability, automation and security platforms over narrow agent-only products if enterprise deployment expands but remains constrained by visibility and integration. Reassess if hyperscalers bundle comparable controls or standalone vendors fail to show attach-rate or renewal improvement.
• Monitor cloud optimization commentary alongside agent adoption. A rise in agent usage without durable workload growth could initially support consumption, then pressure cloud spending as customers impose quotas and controls; cut any cloud-usage thesis if optimization accelerates or consumption guidance weakens.
• Treat Redwood’s reported customer advantage as a sales claim, not proof of product causality. Seek independent evidence—customer wins, renewal/expansion metrics, or disclosed agent-governance revenue—before taking company-specific exposure.