As enterprises expand their use of autonomous AI agents, rising demand for data processing and infrastructure is making cost control a more immediate concern for technology leaders. An Ocient report, based on a survey of more than 500 U.S. IT and data leaders conducted with market research firm Propeller Insights, found that three-quarters of respondents said their costs are increasing.
The findings point to a widening gap between AI ambitions and the resources available to support them. About one-quarter of respondents said AI demand has outgrown their existing infrastructure, while fewer than one-quarter said they could support growing AI workloads without significant changes. Unpredictable costs were identified as a leading data challenge by 41% of executives, up from 35% two years earlier.
Agent workloads add pressure to infrastructure
Agentic systems can generate more queries and require additional data processing, while also raising expectations for low-latency responses. Those requirements can increase the burden on analytics systems and underlying infrastructure, according to the Ocient report. For organizations, the challenge is not simply whether to use AI, but how to account for workload growth and expense as deployments develop.
The cost pressure arrives as companies continue to pursue returns from AI investments. Gartner data cited in the report indicates global AI spending is set to nearly double in 2026, partly because of rising infrastructure demand. That growth makes it important for technology leaders to distinguish between initiatives delivering value and those consuming resources without sufficient results.
Data readiness is another constraint. A recent Dun & Bradstreet survey found that only 8% of enterprises considered their data ready for AI. The Ocient report notes that poor data readiness, rather than limitations in the technology itself, has contributed to rising project failure rates in recent years. Investments in AI capacity may therefore fail to deliver expected results if organizations have not also addressed the data foundations those systems depend on.
Funding decisions shift from expansion to prioritization
Only 38% of leaders surveyed by Ocient said they had seen significant return on investment from their AI investments. Nearly one-quarter said they had reduced or paused initiatives that were underperforming, while about 45% said they had increased investment in initiatives they considered successful. The report describes this as a process of cutting what is not working, supporting what is, and refocusing attention on the data beneath AI applications.
These decisions require more than comparing project budgets. Respondents gave nearly equal weight to security risk, financial return, accuracy and cost at scale when deciding which AI projects to continue. That balance reflects the difficulty of evaluating AI deployments on short-term savings alone: a project may have potential, but still require controls or investment to operate safely and reliably as usage grows.
Technology leaders are responding with measures that include additional security or governance requirements, as well as controls over AI usage and spending. Such measures can help organizations track resource consumption and set boundaries around deployment. The survey findings do not establish that any single control will reduce costs, but they show that cost oversight is being considered alongside governance and risk as AI use expands.
Efficiency remains a practical measure of value
In the Ocient survey, leaders reported seeing the most value from AI in operational efficiency, rather than employee productivity. That finding is broadly consistent with a recent SAP report, which said AI was helping employees create insights, make decisions and interact with customers more than it was directly driving cost efficiency or productivity gains.
For cloud and IT teams, the results underscore the need to evaluate workloads continuously rather than treating AI funding as an all-or-nothing decision. Monitoring demand, processing requirements and spending can inform choices about where to scale capacity, introduce stricter controls or pause an initiative. The central task is to connect infrastructure costs to outcomes, while accounting for data readiness, security and performance requirements.
The report frames pruning as a way to make room for investments with stronger prospects, rather than as a retreat from AI. With unpredictable costs rising and many leaders reporting infrastructure constraints, the ability to stop or reshape underperforming work is becoming part of managing AI adoption. The survey suggests that enterprises are increasingly weighing project-level returns against the resources and safeguards needed to sustain them.
