Gartner: Only 28% of AI Infrastructure Projects Fully Deliver ROI
Summary
- • • Gartner study of 783 I&O leaders finds only 28% of AI use cases fully meet ROI expectations.
- • • 20% of AI initiatives fail outright; 57% of I&O leaders report at least one project failure.
- • • ROI is driven by integration, governance, and operational alignment — not model sophistication.
- • • IT service management (ITSM) is the top success area, cited by 53% of I&O leaders with AI wins.
Details
Only 28% of I&O AI use cases fully succeed and meet ROI expectations
Gartner surveyed 783 infrastructure and operations leaders in late 2025. A full 20% of initiatives fail outright, and 57% of respondents reported at least one project failure.
Model sophistication is not the driver of AI ROI — integration and governance are
Gartner's analysis concludes that success depends on how well AI is embedded into existing workflows, how it is governed, and how closely it aligns with real operational needs — not on which model is deployed.
Unrealistic expectations and skills gaps are the leading causes of AI project failure
Teams frequently expect AI to immediately automate complex tasks or fix long-standing issues. When results do not appear quickly, confidence erodes and projects stall before reaching measurable impact.
IT service management (ITSM) is the top AI success domain, cited by 53% of I&O leaders
Cloud operations is also a leading area of AI wins. These structured, process-heavy domains offer the clearest path to measurable AI impact within infrastructure teams.
Gartner recommends treating AI use cases as managed products with cross-functional governance
I&O leaders should work alongside CIOs, security, legal, finance, and data analytics teams using a shared scoring model that evaluates each use case for feasibility, risk, cost, and expected business impact — avoiding duplication and tracking collective outcomes.
Gartner's 28% success rate compares favorably to a 2025 MIT finding that 95% of genAI projects yield no measurable financial return
The improvement likely reflects the more structured, workflow-bound nature of I&O deployments versus broader enterprise experimentation. Still, the majority of I&O AI initiatives are not fully delivering on their stated business cases.
Stat = quantitative survey finding, Insight = analytical conclusion, Research = study-derived finding, Industry Update = sector-level trend, Strategy = recommended action or framework, Context = comparative or background information
What This Means
For enterprise AI practitioners and infrastructure leaders, this data reinforces a pattern that is becoming increasingly difficult to ignore: most AI projects are not paying off, and the reasons are organizational and procedural, not technical. Teams that treat AI as an informal experiment, set aggressive timelines, or skip cross-functional governance are the ones most likely to stall or fail. The practical implication is that I&O teams should concentrate AI investment in domains with well-defined workflows — particularly ITSM — while building the governance scaffolding (executive sponsorship, shared scoring models, product-style management) that Gartner associates with the minority of projects that actually deliver.