Artificial intelligence budgets are rising faster than many organisations can track what they are buying or demonstrate what it delivers, according to a new survey of 1,500 senior executives across five countries. The research found that 79% of organisations increased AI investment over the previous year, with spending up by 31% on average. Yet 71.5% lacked comprehensive, organisation-wide visibility into adoption, usage and outcomes.
The findings sharpen a practical question for business leaders: how can they judge whether a growing portfolio of AI tools is worth its cost when activity is scattered across teams and business units? The survey, commissioned by HTEC and conducted independently by Censuswide in July, covered C-suite executives in the United States, United Kingdom, Germany, Saudi Arabia and the United Arab Emirates.
Spending growth is not the same as measured return
The 79% figure refers to organisations whose AI investment increased over the preceding year; the 31% is the average rise in spending. Neither figure measures the total amount invested, the share of budgets devoted to AI, or the financial return obtained. The results describe reported changes, not audited company accounts, and should not be read as proof that investment produced better performance.
The report suggests that organisations are also changing what they expect from AI. Only 18% of executives ranked cost reduction among their most important outcomes. Decision-making, productivity, customer experience, innovation and faster time to market featured more prominently. This broader set of ambitions may make evaluation harder: a cost-saving target can be expressed in money, while improved decisions or customer experience require agreed measures and a baseline before results can be attributed.
Only 27% said their organisation had clearly communicated AI’s purpose and expected outcomes across the business. Without a shared statement of what a deployment is meant to improve, departments can adopt similar systems for different reasons, report incompatible measures, or mistake usage for value. The study says no single measurement model has become standard; organisations draw on strategic, financial, operational, workforce and customer outcomes.
Visibility and governance gaps
For 71.5% of respondents’ organisations, visibility into AI adoption, usage patterns and outcomes was not comprehensive across the enterprise. The report describes a range of partial arrangements, from function-level tracking and fragmented reporting to anecdotal feedback or no visibility. That is not the same as saying those organisations have no controls at all, but it makes a consolidated account of where systems are used and what they cost more difficult.
Cost tracking matters because roughly a quarter of executives reported higher AI usage costs linked to insufficient tracking. A separate 26% cited increased reliance on external partners as an unexpected consequence of adoption. Those are proportions of executives reporting these outcomes, not a calculation of total financial losses or a claim that every organisation experienced them.
Risk measurement appears less developed still: only 14% said their organisation systematically measured AI risk across its portfolio. The report does not establish that the remaining organisations have no safeguards; rather, it points to inconsistent or non-systematic measurement. As tools gain the ability to take actions rather than merely generate recommendations, organisations need to know which decisions are automated, who remains accountable, when a human review is required and how incidents can be traced.
Different sectors face different obstacles
The overall averages conceal contrasts between industries. In the survey, 52% of telecommunications organisations were either scaling established AI initiatives or described themselves as AI-native, while 28% of healthcare respondents remained at an early exploration stage. These are survey classifications, not independent assessments of the quality or commercial success of deployments.
The challenges also varied. Manufacturing executives often identified data availability as a primary measurement obstacle; automotive respondents pointed to tooling limitations; and retailers highlighted difficulty attributing outcomes to AI. Financial services and telecommunications respondents put greater weight on decision quality, while manufacturing prioritised productivity. That variation argues against a single company-wide scorecard imposed without regard to how each use case is meant to work.
For organisations trying to close the gap, the immediate task is less about producing another headline adoption count than establishing a usable inventory: which systems are in use, which teams own them, what data and external services they consume, and what outcome each deployment is meant to change. A baseline and review period make it possible to compare costs and results, while recording exceptions and human oversight clarifies who is responsible when a system behaves unexpectedly.
The survey offers a snapshot of executive views gathered in July, not a longitudinal test of AI returns. Its figures can help frame questions for boards and operating teams, but they cannot settle whether any particular deployment is effective. The central challenge it identifies is organisational: growth in investment alone does not supply the visibility, shared objectives or risk measures needed to judge whether AI is delivering value.
