EMEA CIOs Rework ROI, Audit Systems to Restart AI Rollouts

IDC research shows European boards are slowing or refocusing AI projects because execution and financial proof have lagged behind technical investment. Over the past 18 months many firms moved AI beyond tests and invested in large language models and machine learning, but only 9% of organisations in the region delivered quantifiable business outcomes from most of their AI projects in the past two years. The other 91% remain in pilots or limited trials.

Procurement teams often compare software costs to expected headcount reductions. IDC finds that approach misses many benefits of generative AI, which can appear as new revenue sources, faster worker output or avoided losses. A predictive maintenance model in a factory, for example, may not reduce engineer headcount but can prevent an assembly-line collapse; the financial value of the avoided incident does not appear on a single departmental spreadsheet.

Without a standard financial framework for these indirect benefits, pilots are judged on narrow metrics and lose funding before reaching production. IDC recommends CIOs recast ROI calculations so AI outcomes map to corporate financial measures and so teams can justify ongoing infrastructure and operations spending, not just initial development costs.

Scaling pilots into live services exposes architecture and data gaps. Innovation budgets can cover prototype API calls and cloud tests, while production deployments require continuous compute, active data pipelines and maintenance. Engineering teams report friction when integrating modern vector databases and retrieval systems with older on-premise Oracle or SAP servers. Retrieval-augmented generation architectures need clean, categorised data; running large models on disorganised storage produces lower-quality outputs and higher rates of hallucination. Restructuring data to address these issues is expensive and time-consuming.

Continuous inference and model tuning raise cloud bills, and finance teams are scrutinising hyperscaler costs. European data protection and cybersecurity rules also set deployment parameters that increase baseline operational expenses. Work to secure models against prompt injection, to log decisions and to document model behaviour adds engineering and compliance workload. Some firms treat these legal requirements as constraints; others use compliance obligations to enforce stronger system architecture early in development and report improved operational resilience, better ESG performance and higher customer trust after building governance from the start.

User adoption is a separate barrier. Engineering teams sometimes deliver tools that do not match daily workflows, and employees then avoid them. Firms that redesign tools around existing tasks, reduce friction and invest in reskilling and training report higher uptake. An automated contract-review system, for instance, must free lawyers for negotiation work rather than simply replace routine checks, or adoption will stall.

IDC also reports that 42% of EMEA C-suite leaders expect the CIO to lead digital and AI transformation with a focus on creating new revenue sources. The research recommends that CIOs connect technical projects to measurable business outcomes, enforce cross-department alignment and update operating models to support continuous AI operations so pilots can move into production.