A Gartner survey published on July 20, 2026, found that finance organizations are pouring most of their AI budgets into productivity while boards are looking for something else entirely. Drawing on responses from 204 finance leaders collected in March 2026, the survey reported that 45 percent of finance AI investments lean toward productivity and efficiency, and only 20 percent lean toward improving the quality of decisions. Gartner frames the split as a widening distance between what finance is buying and what the board expects that spending to produce.
Shankar Keshav, a principal analyst in Gartner's finance practice, said that many CFOs are prioritizing productivity use cases while boards place greater emphasis on investments that "drive growth, improve decision-making and deliver competitive advantage." The consequence, in Gartner's telling, is a perception gap, where finance leaders report genuine progress on adoption and the board sees limited strategic impact. Even a well-run AI project can disappoint when it answers a question the board was not asking.
The finding lands in the middle of a debate GAIG built an entire category around, which is how an organization proves what its AI spending returned. Gartner is describing the same wall the AI ROI field keeps hitting, that hours saved is the easy number to report and the hard one to turn into enterprise value.
Conditions Driving This Change
Finance teams reached for the AI use cases that were simplest to deploy and measure, which tended to be productivity and efficiency gains inside the finance function itself.
Productivity gains carry a ceiling, because once a task becomes faster the benefit plateaus unless that speed changes a broader business decision or lets the function operate in a different way.
Boards have started asking for growth, better decisions, and competitive advantage from AI, rather than a count of pilots or a tally of hours saved.
The metrics most finance teams report, meaning adoption rates and time saved, do not translate cleanly into the enterprise outcomes a board is weighing.
Gartner's data shows that functions investing in initiatives that create new products, markets, or value propositions were more than twice as likely to report high realized value from AI.
Enterprise AI spending has climbed faster than the accounting built to justify it, a gap that independent research outside Gartner has also documented.
The pressure to show a return has moved up to the board level, where the question is strategic impact rather than internal efficiency, and where the current metrics fall short.
What AI ROI Looked Like Before This
For the past two years, the safe move in enterprise AI was to buy productivity. A finance team could license an assistant, point it at a repetitive process, and show a chart of hours saved within a quarter. The gains were real and easy to measure, and they made a clean story for a status update, which is a large part of why so much of the budget went there.
The trouble is that productivity gains flatten out. Speeding up a task returns time to the people doing it, and that return stops growing once the task is about as fast as it can get. Converting that saved time into money requires it to change something larger, a decision, a headcount plan, or a process the business runs, and that second step is the one most programs never took.
The reporting stayed shallow at the same time. Finance told the board how many pilots were live and how many hours the tools had saved, because those were the numbers the tools produced. The board heard activity where it wanted impact, and the distance between the two grew quietly until someone put a figure on it.
What It Looks Like Now
Gartner has now attached numbers to that distance. With 45 percent of finance AI investment aimed at productivity and only 20 percent aimed at decision quality, the survey gives a shape to a complaint that had been anecdotal, and it hands CFOs a benchmark against which to judge their own portfolios. The 20 percent figure is the one that should stop a finance leader, because decision quality is far closer to what a board means by value.
The survey also points to what the higher-returning teams did differently. Functions that invested in what Gartner calls Upend initiatives, the ones that create new value propositions, products, or markets, were more than twice as likely to report high realized value from their AI. Keshav's prescription follows from that, a portfolio approach that shifts money toward decision-making, scenario analysis, growth, and reusable assets such as data, models, and knowledge, along with a change in how success is measured, moving from counting pilots and hours toward enterprise impact.
One caveat belongs on the Upend figure, and it is the caveat this whole category turns on. Reporting high realized value is a self-assessment, so the survey is measuring what teams believe they got back rather than a verified return, which is the same soft spot that undermines most AI ROI claims. The direction of Gartner's advice is sound, and the number behind it carries the asterisk that a self-reported value figure always carries.
Our Take
AI ROI Take
Gartner is describing a measurement problem wearing the costume of a strategy problem. A finance team cannot rebalance toward enterprise value if it cannot measure enterprise value, and the reason so much spending sits in the productivity column is that hours saved is the one number these tools produce without a fight. The board is asking for a figure that most finance functions are not yet instrumented to give.
The fix Gartner points to, measuring enterprise impact instead of pilots and hours, is the right target and the harder one, because it means connecting AI cost to the decisions and outcomes it influenced rather than to the time it returned. That work runs through the same layers any credible AI return figure requires, from seeing the full cost of a use case to attributing it to the outcome it changed. A portfolio shift toward decision quality only reads as value if the measurement behind it can survive a board's second question.
The survey is one more sign that the market is moving from counting AI activity toward proving AI value, which is the shift GAIG built its AI ROI category to track. Finance leaders working out how to measure enterprise impact rather than hours saved can start with the four-layer framework in our AI ROI metrics guide, and compare the platforms that own each measurement layer in the AI ROI category at GetAIGovernance.net.
