Study Finds Most Organizations Are Not Consistently Measuring AI Value Against Costs
While AI investment continues to grow, relatively few organizations consistently measure whether the technology's benefits justify its costs, according to consulting firm KPMG.
The company's Global AI Pulse Q3 2026 report found that 89% of organizations report measurable value from AI, while just 12% consistently assess that value against costs across the organization.
Average planned AI investment for the next 12 months has risen to $210 million, compared with $186 million in the first quarter of 2026.
The findings are based on responses from 2,131 senior leaders across 20 countries and jurisdictions, surveyed between July 23 and August 26, 2026.
AI Spending Rises, but Measuring Returns Remains a Challenge
Most organizations have established ways to track AI spending: 61% review costs during project approval, and 59% monitor them once systems are operating.
Consistently connecting those costs to value remains less common. Beyond the 12% that do so across the organization, another 26% assess value against costs for most initiatives.
Organizations reporting established AI returns have stronger measurement practices. Among this group, 48% consistently assess value against costs, four times the proportion in the overall survey group.
Cost-monitoring dashboards are also more common among organizations reporting established returns: 77% use them, compared with 43% of those still experimenting.
"The organizations seeing a return did the implementation and the transformation properly: guardrails in place, workforce ready, data cleaned up," said Priya Emmanuel, global head of aIQ at KPMG International. "That is a sequence, not a secret."
AI Governance and Accountability Gain Attention
Organizations are also developing more formal systems to manage AI deployment.
Fifty-five percent operate a formal AI management layer, which KPMG calls a "harness layer," comprising controls and tools between AI models and business applications.
Security and identity controls, along with data access controls, were the most common capabilities, each reported by 43% of respondents. Model routing, which determines which model handles a particular task, was reported by 23%.
Meanwhile, 53% assign accountability for AI-informed decisions to a named executive, the CEO, or an executive committee.
The proportion reporting significant employee adoption of AI agents rose from 25% in the first quarter to 34% in the third quarter.
Cybersecurity Concerns Increase with AI Maturity
Cybersecurity concerns were more common among organizations with mature AI programs.
Half of those reporting established AI returns identified cybersecurity as a barrier, compared with roughly one-third of those still experimenting.
Overall, 86% of respondents said their organizations are adapting their cybersecurity operating models to address AI-accelerated threats.
The report identifies accountability, operational resilience, and cost management as priorities as AI becomes more embedded in business operations.
"Many organizations have made AI models widely available and track their use," said Swaminathan Chandrasekaran, head of the US AI Center of Excellence at KPMG. "But consumption alone is not a strategy."
The full report is available here on the KPMG site.
Featured
-
Nvidia Releases Nemotron 3.5 Lightning Open AI Model
Nvidia has introduced Nemotron 3.5 Lightning, a new open AI model designed less as an all-purpose answer engine than as a fast workhorse inside long-running AI agent systems.
-
Anthropic Adding Invisible Watermarks to Claude-Generated Text
Anthropic recently announced it is introducing machine-readable watermarks to text generated by supported Claude models as it moves to comply with European Union AI transparency rules.
-
Gartner Marks Enterprise Move from AI Experiments to AI Engineering
According to Gartner's recent "Hype Cycle for Enterprise Architecture, 2026" report, enterprises are moving from AI experimentation toward a more industrialized approach to AI delivery.
-
Claude Sonnet 5.5 Offers Model-Switching Cybersecurity Safeguards
Anthropic's latest Claude release, Sonnet 5.5, brings cybersecurity safeguards that can change which model handles a request during a Claude conversation. Higher-risk tasks can fall back to Sonnet 5.