Nearly six in ten organisations reported higher-than-expected AI expenditure over the past year, prompting businesses to tighten spending controls and explore multi-model AI strategies, according to a new report by the Capgemini Research Institute.
The report, based on a September 2026 survey of 500 senior executives across 10 countries and 13 industries, found that 59% of organisations spent significantly more on AI than expected or budgeted. Among affected organisations, expenditure was 24% higher than expected on average, while 21% of surveyed organisations exceeded their AI budgets by more than 30%.
The findings highlight gaps in how businesses estimate the cost of AI deployment. Apart from direct expenses such as model usage and tokens, organisations also need to account for indirect costs, including human oversight, correcting AI-generated errors and governance. Half of the executives surveyed said AI spending was approved without adequate visibility into its full cost. Only 56% were confident in forecasting how AI costs would change as adoption scales.
The rising expenditure is prompting businesses to reconsider their AI strategies. More than three in five executives expect their organisations to adopt a multi-model strategy over the next 12 months, selecting different AI models based on cost, performance, speed and business requirements. Nearly three in five organisations also plan to expand their use of open-weight models and smaller, specialised models for specific tasks.
Volker Darius, CEO, Capgemini Invent, said Cost discipline alone will not create a profitable AI estate, just as value creation without cost control will not be sustainable. Organisations need to measure cost, quality and business value together, and use the most appropriate model for each task.
The report also highlights difficulties in measuring returns from AI investments. While organisations can track how much they spend on AI, many struggle to determine whether these investments are translating into measurable business benefits, such as improved productivity or lower operating costs. This makes it harder to justify further investment, with 57% of executives identifying the demonstration of business value as a major challenge in securing funding for AI initiatives.
Only one in five organisations currently calculates AI Yield, a metric that measures the verified business value generated by AI after accounting for the full cost of delivering an outcome safely. Employee productivity, lower operating costs and faster business processes were among the most commonly reported benefits.
The findings points to a shift in how businesses assess AI investments. While 60% of organisations currently measure total AI expenditure against budget, only 44% expect to prioritise this metric over the next six to 12 months. First-pass success rate, which measures how often AI completes a task correctly on its first attempt, is expected to become the leading performance metric, followed by cost per completed task or workflow.