45% of CFOs Spend AI Budget on Efficiency. Boards Want Growth.

Gartner’s March 2026 survey of 204 finance leaders surfaces a gap between how CFOs spend AI budget and what their boards actually want to see: 45% of respondents say their AI investment leans toward productivity and efficiency, while only 20% say it leans toward improving decision quality. Boards, Gartner notes, tend to weight AI investments that drive growth and sharpen decision-making more heavily than investments that simply trim cost.

The survey’s sharper finding is about payoff, not allocation. Functions that invested in what Gartner calls “Upend” initiatives — projects creating new value propositions, products, or markets, rather than optimizing existing ones — were more than twice as likely to report high realized value from AI as functions that stuck to conservative, productivity-only spend. Gartner’s recommendation is a portfolio approach: deliberately shift more budget toward decision-support use cases, scenario analysis, and reusable assets like data and models, instead of concentrating almost entirely on efficiency plays.

That recommendation lines up with the measurement problem OpenAI CFO Sarah Friar raised days earlier in a widely read essay: that cost-per-token is the wrong yardstick for AI ROI, and the better question is “Useful Intelligence per Dollar” — how much useful work gets done, at what true cost per successful task, with what dependability, and whether returns compound as usage scales. Friar’s framework and Gartner’s survey attack the same blind spot from different sides — one supplies the metric, the other shows what happens when companies default to the wrong one.

For sales and consulting conversations with finance leaders, the two pieces together make a specific pitch easier to land: efficiency-only AI spend is the safe, board-underweighted choice, and the data now shows the growth-oriented alternative pays off more than twice as often.