Gartner finds only 28% of AI projects deliver ROI as most fail to deliver results

Daniel Levi Posted On April 7, 2026

196 Views

The promise of AI has been loud. The results, less so.

A new report from Gartner puts a number to what many teams are starting to feel: most AI projects aren’t paying off. In infrastructure and operations, just 28% of AI initiatives meet ROI expectations. One in five fails outright.

That gap between ambition and outcome is starting to reshape how companies think about AI. The era of experimentation is giving way to a more grounded phase, where results matter more than prototypes.

“Only 28% of AI use cases in infrastructure and operations (I&O) fully succeed and meet ROI expectations, while 20% fail outright, according to a Gartner, Inc. survey of 782 I&O leaders in November and December 2025,” Gartners notes.

The data comes from a survey of 782 infrastructure and operations leaders conducted late last year. What stands out is not just the failure rate, but the pattern behind it. Many projects don’t collapse due to a lack of effort or funding. They stall after expectations run ahead of reality.

Gartner’s latest findings echo earlier research from the Massachusetts Institute of Technology (MIT), which reported in August 2025 that as many as 95% of generative AI pilots failed to deliver meaningful results.

Why Most AI Projects Fail: Gartner Reveals the Real Drivers of ROI Success

Melanie Freeze, Director of Research at Gartner, points to a common mistake: teams expect AI to fix deep operational problems almost immediately. When that doesn’t happen, confidence drops, and projects lose momentum.

“The 20% failure rate is largely driven by AI initiatives that are either overly ambitious or poorly scoped. AI that doesn’t fit into the organization’s operations simply can’t deliver ROI.”

That mismatch shows up most clearly in areas where expectations are highest. Auto-remediation systems, self-healing infrastructure, and agent-led workflows are among the most common failure points. These are complex environments where edge cases matter, and reliability is non-negotiable. AI struggles when pushed beyond what current tools can consistently handle.

There’s another layer to the problem. Skills and data. Around 38% of leaders who reported setbacks said their teams lacked the expertise needed to execute. The same percentage pointed to poor data quality or limited data access. Without clean, usable data, even the best models struggle to deliver meaningful outcomes.

Yet the report doesn’t paint a bleak picture across the board. Many organizations are finding success, and their approach looks very different from the ones that fail.

The biggest shift is how AI is treated inside the organization. High-performing teams don’t run AI as a side experiment. They weave it into the systems people already use every day. That makes adoption easier and results more visible.

There’s a strong leadership component as well. Teams with executive backing move faster and face fewer internal roadblocks. When leadership is aligned, AI projects stay funded and focused.

The third factor is discipline at the start. Teams that succeed tend to begin with clear business cases and realistic expectations. They focus on use cases with proven value, especially in areas like IT service management and cloud operations. These are environments where processes are well defined, and the return is easier to measure.

“High-performing I&O leaders start with realistic AI business cases and upfront preparation.”

That grounded approach is becoming more important as spending ramps up. AI infrastructure is expected to account for more than half of global IT spending this year. That level of investment is drawing more scrutiny from CEOs and CFOs, who want clear outcomes, not just technical progress.

Freeze suggests treating AI use cases like products, with clear ownership, measurable impact, and shared evaluation criteria across teams. This makes it easier to prioritize what gets funded and what doesn’t.

There’s a broader shift happening here. Companies are moving away from asking what AI could do and focusing on what it should do inside their operations. That sounds subtle, but it changes everything.

The takeaway from Gartner’s findings is simple. AI doesn’t fail due to a lack of potential. It fails when it’s disconnected from the business it’s supposed to serve.

The hype isn’t disappearing. It’s being tested. And for many teams, that test is still ongoing.