EY warns businesses risk losing AI productivity gains to complexity

Artificial intelligence is boosting productivity at companies worldwide, but many chief executives are struggling to turn efficiency gains into higher profits and sustainable growth, according to an EY-Parthenon survey that highlights the risks of rising complexity, skills shortages and inadequate investment in workforce development.

The latest EY-Parthenon CEO Outlook survey found that almost all chief executives reported productivity gains over the past 12 months, measured by the ability to generate more output per employee, while nearly half said they were reinvesting those gains in growth, innovation or transformation.

However, close to a quarter of respondents said the benefits were being absorbed by operational complexity, regulatory obligations, risk management requirements and additional workloads, raising questions about whether companies can translate technological improvements into lasting business value.

The findings come as executives contend with slowing economic growth, geopolitical uncertainty and rapid technological change, with many increasingly viewing delayed investment as a greater risk than acting amid uncertainty.

Although equity markets remain elevated, credit is available and capital continues to flow into artificial intelligence, companies face difficult decisions about where to invest and how to position themselves for growth in a volatile environment.

EY said the emerging productivity gains would depend not simply on adopting AI, but on whether businesses redesigned their operating models, reorganised work and deliberately redirected the capacity they created.

“The real AI dividend is not simply lower cost but greater and better output,” the report stated, warning that the next productivity divide would be between organisations that reinvested their newly available capacity and those that allowed it to disappear into complexity and inefficient processes.

AI-enabled tools and redesigned workflows were identified as leading productivity drivers, with the strongest results emerging when new technologies were integrated into fundamentally reworked processes rather than added to existing systems.

A chief executive in North America’s wealth and asset management sector described how their company sought to ensure that AI projects delivered measurable business benefits.

“We require every project to have a named business sponsor, rather than being owned by the technology function, with a formal review after 90 days and a genuine option to stop,” the executive said.

“Two of our first five projects were discontinued,” they added. “That discipline is what makes the returns from the remaining projects credible, and it is the step many firms skip.”

What is more, the report highlighted the difficulty of measuring productivity improvements, particularly when faster individual tasks do not necessarily improve a company’s overall financial performance.

While businesses may be able to identify activities completed more quickly, fewer can demonstrate how the time saved changes their operating costs, output, customer experience or cash generation.

Legacy technology remains a major obstacle, but outdated management structures can also prevent companies from benefiting fully from AI, with unnecessary approvals, reporting requirements and handovers preserving inefficiencies in existing workflows.

EY argued that businesses needed to measure outcomes rather than technology adoption, redesign workflows to eliminate redundant processes while preserving human judgement, and decide in advance whether productivity gains would fund expansion, innovation, workforce development or shareholder returns.

The financial case for AI also remains unclear for many companies, with only 16 per cent of CEOs reporting detailed, real-time visibility into AI costs and high confidence in assessing returns on investment.

A further 56 per cent said they had reliable visibility into most costs and reasonable confidence in their returns, while others continued to face significant gaps in understanding the economics of their investments.

The report stressed that businesses needed to account for the full cost of AI, including model fees, cloud infrastructure, data, integration, cybersecurity, governance, process changes and employee training, as well as the expense of running new and legacy systems simultaneously.

“ The pace of adoption will depend not only on what the technology can do, but whether the value created by each token exceeds its cost,” EY-Parthenon chief economist Gregory Daco said.

Without clear financial measures, AI projects could meet their technical targets while weakening business performance as they expanded.

The survey also highlighted the implications for employment, with 4 in 5 CEOs expecting AI to have a greater effect on job roles, skills and the organisation of work than on overall workforce size over the next 3 years.

As AI agents become capable of performing more duties independently, companies could separate tasks from traditional job roles, reorganise workflows and consolidate responsibilities, although the pace and extent of these changes would vary across organisations.

The report suggested that businesses could eventually sustain growth with fewer employees than would have been required in the past, provided their operating models adapted sufficiently quickly to technological advances.

However, skills shortages threaten to slow that transition, with 47 per cent of CEOs saying their organisations were not developing skills quickly enough to keep pace with AI innovation.

A further 72 per cent expected skills shortages to become a greater barrier to growth than access to capital within 3 years.

Reskilling was already among the leading contributors to productivity improvements, cited by 33 per cent of CEOs, yet only 15 per cent said productivity gains typically funded reskilling and capability building.

A chief executive in Europe’s mobility sector said companies needed to explain the consequences of automation more clearly to employees.

“People need to understand specifically what automation will mean for them, and we have not always been good enough at explaining that,” the executive said.

“When you announce automation to a workforce in this industry, people do not hear ‘productivity’. Instead, they hear ‘job losses’,” they added.

The executive added that companies needed to explain earlier which roles would change and how, while providing genuine training pathways and employment opportunities following retraining.

EY said the skills challenge extended beyond hiring AI specialists, requiring businesses to develop domain experts capable of redesigning work, managers able to determine where human oversight remained essential, and employees who could use AI critically.

The report cited the OECD’s 2026 Skills in the AI Age publication and International Monetary Fund research on emerging skills, both of which emphasise the importance of combining foundational, digital and complementary skills with the organisational capacity to deploy them effectively.

EY concluded that companies needed to treat AI as a continuous reinvestment cycle, measuring financial and workforce outcomes together, directing released capacity towards clearly defined priorities and removing processes that could otherwise absorb productivity gains.

Where businesses cannot develop essential capabilities quickly enough, partnerships, acquisitions and other portfolio decisions may be needed to close gaps and support growth.