- Only 28% of AI use cases meet ROI expectations, while 20% fail outright, according to a Gartner survey
- AI maturity stagnated year over year, even as organizations set ambitious two-year targets for AI-augmented network operations, according to a separate IDC study
- Both studies point to the same prescription: embed AI in existing workflows, secure executive buy-in and focus first on use cases with proven value
AI in networking, infrastructure and operations is stalling. Two new research reports paint similar pictures of where AI stands in infrastructure and networking: ambitious targets, persistent barriers and real-world adoption that keeps falling short of the plan.
Only 28% of AI use cases fully succeed and meet ROI expectations. Twenty percent fail outright. The remaining projects land somewhere in between, delivering partial value while leaving organizations uncertain what to try next, according to Gartner, after a survey of 782 enterprise infrastructure and operations (I&O) leaders conducted late last year.
"AI that doesn't fit into the organization's operations simply can't deliver ROI," said Gartner research director Melanie Freeze in a statement on the firm's website. Freeze attributed the 20% failure rate largely to initiatives that were either overly ambitious or poorly scoped. These organizations expected AI to immediately automate complex tasks, cut costs or resolve long-standing operational problems. When those results didn't materialize quickly, confidence eroded and projects stalled.
Thirty-eight percent of I&O leaders who experienced setbacks cited persistent skills gaps. Another 38% pointed to poor data quality or limited data availability as a direct cause of failure.
AI networking maturity stagnates despite high expectations
IDC arrives at parallel conclusions. Despite strong optimism in both 2024 and 2025 surveys, the share of organizations at "Substantial Use" of AI barely moved — holding at roughly 32% heading into 2026 against targets that called for significant forward momentum. IDC's researchers flagged security, complexity, integration challenges and limited visibility as persistent blockers.
IDC's 2026 AI in Networking Special Report is based on a survey of 518 networking professionals conducted in October and November 2025.
The stagnation is notable because expectations have remained high. Respondents in both survey years anticipated major advancement within two years. Instead, the gap between planned and actual progress has become a recurring feature of the data, IDC said.
What separates AI success from failure?
However, not all the news on AI in networking is gloomy. More and more operators are declaring and validating attainment of Level 4 network autonomy in specific domains, according to a recent TM Forum report. And Nvidia sees a surge in AI-driven network automation adoption.
The path forward, both studies suggest, runs through operational discipline rather than technical ambition. Gartner found that successful I&O leaders share three characteristics:
- they embed AI into systems and processes people already use
- they secure full executive backing before deployment
- and they start with realistic business cases grounded in use cases where AI has a track record.
Fifty-three percent of I&O AI wins reported in the Gartner survey occurred in IT service management — a mature domain with clear metrics and well-understood workflows.
IDC's networking data reinforces the point. AI-augmented campus and branch network management tasks jumped from a mean of 20% in the 2024 survey to 31% in 2025, with respondents expecting that figure to reach 54% within two years. Edge AI adoption sits at 27% today, with 54% of organizations planning deployment within two years.
Progress is real — it is just slower and harder than the industry anticipated.
For network operators and communications service providers managing their own infrastructure and operations, the message from both research firms is consistent: the organizations pulling ahead are treating AI as an operational discipline, not a technology project.