Global investment in Artificial Intelligence (AI) is rapidly accelerating, yet a recent report by KPMG's first quarterly Global AI Pulse survey highlights a significant and widening disparity between enterprise AI spending and the achievement of measurable business value. The survey reveals that global organizations plan to invest a weighted average of $186 million in AI over the next 12 months. However, a stark reality is that only 11 percent of these organizations have successfully deployed and scaled AI agents to produce enterprise-wide business outcomes. While 64 percent of respondents indicate that AI is already delivering "meaningful business outcomes," KPMG clarifies that this often refers to incremental productivity gains rather than the compounding operational efficiencies that drive substantial improvements in profit margins.

The report delineates between "AI leaders," defined as organizations actively scaling or operating agentic AI, and all other entities. The performance gap between these two groups is considerable. Steve Chase, Global Head of AI and Digital Innovation at KPMG International, emphasizes that increased AI spending does not automatically equate to value creation. AI leaders, making up the 11 percent, are distinct because they move beyond mere enablement to deploy AI agents that reimagine processes and fundamentally reshape decision-making and workflow across the enterprise. Among these leaders, 82 percent report meaningful business value from AI, a figure that drops to 62 percent for their peers. This 20-percentage-point difference, though seemingly modest, reflects not just superior tools but fundamentally different deployment philosophies. AI leaders utilize agents to coordinate cross-functional work, automate decisions without constant human intervention, generate near real-time enterprise-wide insights from operational data, and proactively flag anomalies. In IT and engineering, 75 percent of AI leaders use agents for code acceleration, compared to 64 percent of their peers. In operations, where supply-chain orchestration is key, the split is 64 percent versus 55 percent, indicating deeper process re-architecture by leaders who redesign processes before deploying agents, in contrast to others who merely layer AI models onto existing workflows.

A closer look at the $186 million average investment reveals interesting regional variances: ASPAC leads at $245 million, followed by the Americas at $178 million, and EMEA at $157 million. Within ASPAC, organizations in China and Hong Kong average $235 million, while US organizations within the Americas average $207 million. These figures encompass spending on model licensing, compute infrastructure, professional services, integration, and essential governance and risk management for responsible AI scaling. The crucial question, however, is the proportion of this investment allocated to the operational infrastructure needed to actually derive value from the models. The survey suggests most organizations underinvest in this latter category. While compute and licensing costs are straightforward to budget, "friction costs" – such as engineering hours for integrating AI with legacy ERP systems, latency from retrieval-augmented generation pipelines built on poorly structured data, and compliance overhead for AI-assisted decisions in regulated industries – often emerge late in deployment cycles and frequently surpass initial estimates. Vector database integration serves as a prime example, demanding significant engineering complexity and ongoing operational costs for building and maintaining infrastructure (e.g., selecting providers like Pinecone, Weaviate, Qdrant, embedding and indexing proprietary data, and managing refresh cycles), which are rarely fully accounted for in initial AI proposals. The absence or poor maintenance of such infrastructure can significantly degrade agent performance, making diagnosis difficult as models perform correctly based on stale or incomplete context.

Perhaps one of the most practical insights from the KPMG survey pertains to the relationship between AI maturity and risk confidence. Only 20 percent of organizations in the AI experimentation phase express confidence in managing AI-related risks, a figure that jumps to 49 percent for AI leaders. While 75 percent of global leaders cite data security, privacy, and risk as ongoing concerns irrespective of maturity, maturity transforms how these concerns are operationalized. This distinction is vital for boards and risk functions that often view AI governance as a constraint. The KPMG data suggests the opposite: robust governance frameworks enable AI adoption among mature organizations. The confidence to deploy agents in higher-stakes workflows and expand agentic coordination across functions directly correlates with the maturity of the governance infrastructure. Organizations treating governance as a retrospective compliance layer face a double disadvantage: slower deployments due to frequent governance reviews for new use cases and increased exposure to operational risks because embedded governance mechanisms are absent, leading to the discovery of edge cases and failure modes in production. Organizations that integrate governance directly into the deployment pipeline—through tools like model cards, automated output monitoring, explainability tooling, and human-in-the-loop escalation for low-confidence decisions—are those operating with the confidence necessary to scale effectively. Steve Chase concludes, "Ultimately, there is no agentic future without trust and no trust without governance that keeps pace." He emphasizes that sustained investment in people, training, and change management is key to responsible AI scaling and value capture.

For multinational corporations managing global AI programs, the KPMG data highlights significant regional divergences in deployment velocity and organizational readiness. ASPAC is leading the charge in agent scaling, with 49 percent of organizations doing so, compared to 46 percent in the Americas and 42 percent in EMEA. ASPAC also excels in orchestrating complex multi-agent systems, with 33 percent engagement. Barriers to deployment also vary regionally. In both ASPAC and EMEA, 24 percent of organizations cite a lack of leadership trust and buy-in as a primary obstacle, while this figure drops to 17 percent in the Americas. This suggests that organizational cultures where decision accountability is highly centralized can generate institutional resistance to agentic systems that make decisions without per-instance human approval, necessitating careful governance design to define agent autonomy, escalation triggers, and accountability. Additionally, the expectation gap regarding human-AI collaboration is crucial for designers of global agent-assisted workflows. East Asian respondents anticipate AI agents leading projects at a rate of 42 percent, while Australian respondents prefer human-directed AI (34 percent), and North American respondents lean towards peer-to-peer human-AI collaboration (31 percent). These cultural differences demand localization complexity in designing agent-assisted processes across various regional deployments.

A notable finding for CFOs and boards is that 74 percent of respondents indicate AI will remain a top investment priority, even amidst a recession. This may signify genuine conviction about AI's role in cost structure and competitive positioning, or it could reflect an untested collective commitment. Regardless, it signals that the window of opportunity for organizations still in the experimentation phase is closing. If the 11 percent of AI leaders continue to compound their advantage—as suggested by the KPMG data—the imperative for the remaining 89 percent is not merely to accelerate AI deployment, but to do so strategically, without exacerbating existing integration debt and governance deficits that are currently hindering their returns.