See also (wiki): workflow-redesign · data-readiness · ai-maturity-models
Executive Summary
- Only 28% of AI use cases in infrastructure and operations (I&O) fully succeed and meet ROI expectations; 20% fail outright (Gartner, n=782 I&O leaders, Nov–Dec 2025).
- The dominant failure mode is not technical — 57% of leaders who failed said they expected too much, too fast. Overambition kills more AI projects than poor models.
- Success correlates with integration into existing workflows (33% of winners), executive sponsorship (26%), and cross-functional collaboration (25%) — not model sophistication.
- 53% of AI wins in I&O occur in IT service management (ITSM), where the use cases are mature and bounded. Auto-remediation, self-healing infrastructure, and agent-led workflows are where failures concentrate.
- 38% of leaders who faced setbacks cite persistent skills gaps; another 38% cite poor data quality or limited data availability as a direct cause of failure.
The Pattern: Workflow Integration Beats Model Sophistication
Gartner’s April 2026 survey of 782 I&O leaders reinforces a pattern visible across every major study of enterprise AI deployment: the technology is not the bottleneck. Execution is.
The 28% success rate for I&O AI use cases sits squarely between BCG’s finding that only 26% of companies generate significant value from AI (n=1,800, Sep 2024) and McKinsey’s finding that just 6% of organizations are “high performers” capturing 65%+ of AI’s potential value (n=1,993, Nov 2025). The convergence across independent surveys is striking — roughly one in four organizations gets this right.
What separates the 28% from the rest is not better tools. Gartner identified three success factors, and none involve model selection or infrastructure scale:
Embed AI into existing workflows. 33% of successful I&O leaders attribute their wins to integrating AI into the systems and processes people already use. This echoes McKinsey’s finding that workflow redesign is the #1 EBIT predictor out of 25 attributes tested (Mar 2025, n=1,491). AI deployed as a side project — a separate tool, a new dashboard, an isolated pilot — reliably fails.
Secure executive sponsorship. 26% of successful leaders reported full executive support; 25% cited cross-functional collaboration. AI projects that lack C-suite air cover lose funding at the first quarter of disappointing results. Executive backing keeps the investment funded through the adoption curve.
Start with bounded, proven use cases. 53% of I&O AI wins occur in ITSM — ticket classification, knowledge retrieval, automated responses to common requests. These are mature, well-defined processes with clear success metrics. The failures concentrate in ambitious categories: auto-remediation, self-healing infrastructure, and agent-led cross-system workflows. These demand a level of reliability that current AI tools cannot consistently deliver in complex, unpredictable IT environments.
Where Failures Concentrate
The 20% outright failure rate is notable, but the more actionable number is the 57% of I&O leaders who reported at least one AI failure. Their diagnosis: they expected too much, too fast.
This is the same pattern Atlan documented across 200 enterprise AI deployments — organizations that skip workflow redesign and jump to advanced use cases capture a fraction of the available value. The Gartner data adds specificity: auto-remediation, self-healing infrastructure, and agent-led management of workflows within and between systems are the use cases most likely to fail. These require AI to operate autonomously in high-stakes, unpredictable environments — precisely the conditions where current models are least reliable.
Two persistent blockers appear in nearly equal measure among leaders who faced setbacks:
- 38% cite skills gaps. The teams deploying AI lack the expertise to scope, implement, and maintain AI-augmented workflows.
- 38% cite data quality or availability. AI applied to messy, incomplete, or siloed operational data produces unreliable outputs.
Both blockers are organizational, not technological. No model upgrade fixes a skills gap or a data quality problem.
Key Data Points
| Finding | Value | Source | Date |
|---|---|---|---|
| I&O AI use cases that fully succeed and meet ROI | 28% | Gartner (n=782) | Apr 2026 |
| I&O AI use cases that fail outright | 20% | Gartner (n=782) | Apr 2026 |
| I&O leaders reporting at least one AI success | 77% | Gartner (n=782) | Apr 2026 |
| I&O leaders reporting at least one AI failure | 57% | Gartner (n=782) | Apr 2026 |
| Failure attributed to expecting too much, too fast | 57% of those who failed | Gartner (n=782) | Apr 2026 |
| Success attributed to workflow integration | 33% of successful leaders | Gartner (n=782) | Apr 2026 |
| Success attributed to executive support | 26% of successful leaders | Gartner (n=782) | Apr 2026 |
| Success attributed to cross-functional collaboration | 25% of successful leaders | Gartner (n=782) | Apr 2026 |
| AI wins occurring in ITSM | 53% | Gartner (n=782) | Apr 2026 |
| Skills gaps as cause of AI failure | 38% of those with setbacks | Gartner (n=782) | Apr 2026 |
| Data quality/availability as cause of failure | 38% of those with setbacks | Gartner (n=782) | Apr 2026 |
| AI infrastructure as share of global IT spending (2026) | 54% | Gartner | Apr 2026 |
What This Means for Your Organization
The Gartner data confirms what the best evidence from BCG, McKinsey, and Atlan already indicates: the organizations capturing AI value are not running more ambitious projects — they are running better-scoped ones, embedded in existing workflows, with executive sponsorship and realistic timelines.
For a CIO or CTO at a mid-market company, the practical implication is clear. Start with ITSM. Ticket classification, knowledge retrieval, and automated responses to routine requests are where 53% of I&O leaders report their wins. These use cases have bounded scope, measurable outcomes, and mature vendor tooling. Auto-remediation and self-healing infrastructure sound compelling in a vendor demo, but they are where failures concentrate — save them for after the organization has built the muscle of integrating AI into daily operations.
The 38% skills gap and 38% data quality findings are a diagnostic checklist. Before approving the next AI project, ask two questions: Does the team have the skills to scope and maintain this? Is the underlying data clean enough to produce reliable outputs? If the answer to either is no, the project will stall — regardless of how good the model is.
If these findings raise questions about how to sequence AI investments in your infrastructure organization, I’d welcome the conversation — brandon@brandonsneider.com.
Sources
-
Gartner, “AI Projects in I&O Stall Ahead of Meaningful ROI Returns,” Q&A with Melanie Freeze, April 7, 2026. Survey of 782 I&O leaders, Nov–Dec 2025. Credibility: HIGH — Gartner is an independent analyst firm; survey methodology is standard for their press releases; n=782 is a strong sample for I&O-specific research. URL: https://www.gartner.com/en/newsroom/press-releases/2026-04-07-gartner-says-artificial-intelligence-projects-in-infrastructure-and-operations-stall-ahead-of-meaningful-roi-returns
-
Cross-references: BCG “AI Radar 2025” (n=1,800, 26% generate significant AI value); McKinsey “State of AI” Nov 2025 (n=1,993, 6% high performers); Atlan 200-deployment analysis (workflow redesign as prerequisite); McKinsey Mar 2025 (n=1,491, workflow redesign #1 EBIT predictor).
Brandon Sneider | brandon@brandonsneider.com April 2026