A global study by EY and Oxford Economics shows that most companies still rely on pilot projects and isolated artificial intelligence solutions rather than scaling AI across the organisation. The research surveyed 2,500 technology leaders in 28 countries.
Experimentation outpaces deployment — the "AI ROI trap"
The report highlights that while investments in generative and agentic AI remain substantial, many organisations see expected business benefits arrive slower, in smaller amounts, or unevenly. In many cases the pace of experimentation exceeds the speed at which companies can deploy, operate and measure solutions, a phenomenon the study calls the "AI ROI trap." This gap between testing and measurable outcomes is limiting return on investment for numerous respondents.
Governance and measurement gaps
Although more than half of respondents have processes to assess readiness for AI adoption, these processes are often applied inconsistently. Only 33 percent of respondents regularly check whether their IT infrastructure adequately supports AI systems, and just one in four organisations (25 percent) continuously monitors data quality and reliability. These shortfalls directly affect the ability of companies to measure and demonstrate the value created by AI.
What determines long-term success?
According to the report, long-term success depends on whether companies can use AI to transform end-to-end business processes, link AI initiatives to clear business objectives, and measure outcomes with appropriate performance metrics. Erik Slooten, partner on EY AI Confidence, said: “Investments in artificial intelligence alone are not a guarantee of success. Alongside testing technology, deliberate strategic planning, well-defined performance metrics and consistent governance frameworks are needed. In our experience, the companies that extract the most value from AI are those that do not think in terms of isolated solutions, but can scale successful initiatives at the enterprise level.”
About the research
The study is based on two international surveys conducted with Oxford Economics. The first online survey included 1,500 technology industry leaders from 28 countries across the Americas, Asia and the Pacific, and the EMEIA region. A second survey added another 1,000 similar respondents. The research covered hardware, software and SaaS, internet and social commerce, and IT services and cloud companies, and focused on organisations that already have an enterprise-level AI strategy or concrete plans to implement one. It examined AI use cases, governance frameworks, data strategy, investment approaches and enterprise-level value creation.
Conclusions
The report warns that under current practices many companies face an "AI ROI trap": experimentation and investment exist, but inconsistent deployment, operation and measurement may prevent expected returns. It recommends that firms adopt strategic planning, consistent governance and clear performance metrics to make the value created by AI measurable and pursue scalable, enterprise-level solutions.
