A commissioned study of 1,000 senior technology and data leaders, conducted by Wakefield Research on behalf of Teradata, puts a number to the gap: just 7% of global enterprises have reached what the report calls the Operationalizing stage, where AI agents execute multi-step workflows with measurable business impact. The majority — 68% — remain in Experimenting or Developing, still waiting for pilots to become products.
The report, titled "Arrested Automation: Why Agentic AI Stalls at the Enterprise Level," argues that the fundamental problem is not ambition or investment. Nine in ten senior technology leaders expect to increase their agentic AI budgets over the next twelve months; 63% say they have seen no more than a small or emerging return on what they have already put in.
The research identifies context fragmentation as the core obstacle: enterprise data that exists but carries no usable meaning for AI agents. According to the survey, 77% of executives say that 20% or less of their enterprise data is sufficiently contextualised for agents to act on it. A separate 78% report difficulty unifying data and knowledge across business functions so agents can reason across the full organisation.
The distinction between personal AI and organisational AI runs through the report's argument. Tools that help individuals work faster — writing assistants, code helpers — deliver real productivity gains that do not show up at the P&L level. Organisational AI, which executes decisions on behalf of the whole company using shared knowledge and governance, is where the enterprise ROI figures lie. Most organisations are measuring their AI investments against the personal tier while trying to justify infrastructure spend at the organisational one.
Four in ten pilots never reach production: 40% of technology leaders report that more than 40% of their AI pilot projects fail to move beyond the experimental phase, with data infrastructure built for human users rather than autonomous agents cited as the primary cause. Only 15% of organisations successfully get 80% or more of AI pilots into production.
A perception divide between organisational levels adds a further complication. The study found that 69% of C-suite executives believe their organisation is already operating with agentic AI, against 57% of VPs. The gap is not trivial: when leadership believes a transition has already happened, the pressure to invest in foundational data work diminishes.
Teradata frames the path forward around what it calls Autonomous Knowledge — enterprise data equipped with the context, lineage, and governance that agents need to act reliably at scale. The company's prescription is selective rather than comprehensive: audit the highest-value data, embed governance at the data layer, and build for portability across cloud and on-premises environments, rather than attempting to contextualise the entire data estate at once.
The survey covered 1,000 leaders at VP level or above at companies with at least 500 employees, across the US, UK, France, Germany, Japan, and Saudi Arabia. Fieldwork ran from 23 March to 5 April 2026. The full report is available at teradata.com.
To stay across the latest in cloud, AI and enterprise tech analysis from Compare the Cloud, subscribe to our weekly newsletter at https://www.comparethecloud.net/newsletter
