Enterprise AI Adoption: Pilots, Production and ROI — 2026-09-18

Top developments

Nearly half of CIOs struggle to fund over-budget AI initiatives

A new survey by the Futurum Group of 1,636 global enterprise technology decision-makers found that 47% lead organizations that are over budget on their artificial intelligence spending, compared to just 32% who report spending in line with plans. This financial strain occurs as AI becomes increasingly embedded across products and features, often unintentionally inflating costs beyond initial pilot budgets. The findings highlight a critical disconnect between executive expectations for rapid returns and the reality of escalating infrastructure and integration costs.

Only 6% of companies get value from AI despite high adoption

Forbes reports that only 6% of companies are currently getting significant value from their AI investments, signaling a severe lag between adoption metrics and business impact. The article notes a shift in buyer behavior from relying on single models to adopting mixed-model strategies coupled with robust governance frameworks. This low yield rate underscores why many enterprises are re-evaluating their generative AI portfolios and moving away from broad experimentation toward targeted, high-value use cases.

UiPath survey reveals orchestration is key to AI agent ROI

UiPath’s global survey of 590 executives and IT practitioners from large enterprises (>$1B revenue) found that while many have moved past Proof of Concept (PoC), only 31% have successfully deployed AI agents across the enterprise. The study concludes that effective orchestration—the ability to coordinate multiple agents and systems—is the primary determinant of ROI, rather than the sophistication of individual models. This aligns with broader industry observations that most agentic AI pilots stall due to workflow and data integration failures rather than technical model limitations.

AI becomes the primary driver of new cybersecurity spending

AI has emerged as the leading driver of new cybersecurity investments, according to the 2026 Security Budget Benchmark Report by IANS and Artico Search. While overall security budgets grew modestly (5%), new dollars are being disproportionately allocated to AI security demands, such as securing shadow AI usage and protecting model integrity. This trend indicates that as AI adoption scales, security teams are being forced to prioritize AI-specific risk management over traditional perimeter defenses.

Local view

South Korea: Local media reports indicate a surge in AI adoption rates among Korean firms, rising from 48% to 58% in one year, with six out of ten companies now using AI in operations. However, readiness for next-generation technologies like agentic AI remains low at 24%, highlighting a gap between current usage and future capabilities. A Kaspersky survey cited by local outlets found that 41% of Korean companies increased their information security budgets specifically due to AI adoption, driven by concerns over "shadow AI" data leaks. Additionally, KT reported that nearly 60% of financial institutions have moved beyond PoCs to actual implementation, whereas public sector adoption lags at 30%.

Japan: In Japan, a survey by PwC Japan Group highlights that while generative AI is becoming a survival condition for businesses, only 24% of marketers feel it is contributing significantly to business results. Another study by Avitas Education Research Institute revealed that while 54.6% of employees use company-managed AI environments, only 15.1% feel rules are fully embedded, leaving a significant governance gap.

Context & numbers

• Global Spending: Global AI spend is projected to nearly double in 2026, driven primarily by infrastructure demand rather than intentional strategic shifts, as AI becomes embedded in existing software features.
• Abandonment Rates: Gartner continues to predict high abandonment rates, with previous data suggesting 30% of generative AI projects are abandoned after proof of concept, though newer reports suggest up to 89% of agent pilots fail to reach production.
• ROI Disparity: While some firms report a 3.7x average return per $1 invested in generative AI, IBM’s CEO study finds that only 25% of AI initiatives delivered expected ROI, illustrating the variance in outcomes.
• Infrastructure Costs: Hyperscalers are expected to spend more than $1 trillion on data centers next year, raising questions about the sustainability of the current infrastructure boom.

On the radar

• Token Usage as Diagnostic Tool: Forbes contributors are advocating for token usage monitoring not just as a cost-control measure, but as a diagnostic tool to optimize AI operations and identify inefficiencies in workflow integration.
• Shadow AI Governance: With 41% of companies increasing security budgets due to AI, expect a rise in specialized tools for detecting and governing unauthorized "shadow AI" usage within enterprise networks.
• Orchestration Platforms: Following UiPath’s findings, look for increased M&A activity or product launches focused on multi-agent orchestration layers, as enterprises seek to solve the "last mile" problem of scaling pilots to production.