The budget is accelerating faster than the quarter can explain
KPMG's survey of 2,131 senior leaders across 20 countries puts a number on something most enterprise technology leaders have felt anecdotally all year: AI spending is accelerating within the budget cycle itself, not merely growing year over year the way most technology spend does. Average AI investment climbed to 210 million dollars from 186 million dollars in a single quarter, a 13% jump that would be a major line-item surprise in almost any other technology category and has instead become routine inside AI budgets specifically.
That pace of increase mid-cycle is the detail worth sitting with longer than the headline number. Budgets that grow 13% between quarterly reviews are budgets that were under-forecast to begin with, which usually means either the initial estimate badly misjudged genuine organizational appetite, or spending is being approved reactively, deal by deal and use case by use case, rather than against a disciplined annual plan built with this kind of growth rate in mind from the start.
The 12% number that should worry every CFO in the room
Only 12% of organizations consistently measure AI value against cost across their organization, a figure that sits in stark contrast to the budget growth above it. Spending is scaling at double-digit rates per quarter while the discipline to know whether that spending is working remains the exception most organizations have not built yet. This is not a data availability problem in most cases, usage metrics, token consumption, and deployment counts are generally easy to pull. It is a measurement framework problem: knowing what good looks like, agreeing on it across finance and the business unit actually running the use case, and then holding the organization to checking it on a consistent cadence rather than only when a renewal or a board question forces the issue.
The gap between spend and measurement discipline is exactly the kind of finding that tends to surface loudly in next year's budget cycle, once finance teams start asking harder questions about renewal decisions for the AI tools and platforms approved somewhat reactively this year. Enterprise leaders who get ahead of that question now, by building the measurement framework before finance demands it, will be negotiating from a position of data rather than defending a budget line with anecdotes.
Governance works, when anyone actually builds it
The survey's most actionable finding is the gap between organizations with a formal AI harness layer, KPMG's term for structured governance controls spanning approval workflows, monitoring, and accountability, and those without one. Only 55% of organizations overall report having this in place. Among organizations that have already realized measurable AI returns, that figure jumps to 86%, a correlation too large and too consistent to dismiss as coincidence, even without a formal causal study behind it.
That correlation gives CIOs and CFOs a genuinely useful argument heading into next year's planning cycle: governance structure is not the compliance tax that slows AI deployment down, which is how it still gets framed in plenty of internal debates, it is a leading indicator of the organizations actually capturing value from their investment. The 53% of organizations that now place AI accountability at the C-suite level or above reinforces the same point from a different angle, since that kind of senior ownership typically does not exist without the governance structure underneath it to make that accountability mean something concrete.
The regional race is tightening, not widening
KPMG's regional breakdown complicates the narrative that American enterprises hold a commanding, widening AI lead. The Americas does lead at 64% of organizations scaling AI deployment or beyond, but Asia Pacific sits close behind at 61% and EMEA at 56%, and the gap between the fastest and slowest region has narrowed from 16 percentage points to just 8 within the survey's tracked period. A race that looked like it might be settling into a clear leader a year ago is instead compressing toward a three-way tie at the frontier of enterprise deployment.
For multinational enterprises managing AI strategy across regions, that compression argues against treating any single region as the template to export everywhere else. A deployment playbook built around Americas-specific assumptions, lighter regulatory friction, higher risk tolerance, deeper existing cloud relationships, among them, may already be out of step with how fast AI capability and governance maturity are converging in EMEA and Asia Pacific, where survey respondents report nearly identical scaling momentum despite often starting from different regulatory and infrastructure baselines.
Security spend is now inseparable from AI spend
Two final data points tie this survey directly back to enterprise risk planning. 86% of organizations are adapting cybersecurity operations specifically for AI-related threats, and 72% now formally factor model sovereignty, meaning control over where a model runs, what data touches it, and under which jurisdiction, into deployment decisions rather than leaving it as an afterthought once a vendor is already selected. Together these numbers describe an enterprise AI conversation that has fully merged with the enterprise security and compliance conversation, no longer two separate workstreams competing for separate budget lines and separate executive sponsors.
KPMG's Lisa Heneghan frames the overall finding succinctly: what scale demands is the accountability and coordination to run AI well, beyond simply deploying more of it. The data backs that framing specifically. The organizations pulling ahead in this survey are not the ones spending the most, budget growth alone shows almost no correlation with the realized-returns group. They are the ones that built governance, measurement, and security into the deployment from the outset rather than treating those as cleanup work for later, after the spending had already scaled past the point where retrofitting discipline is easy.
