What CFO surveys say about measuring AI ROI: Gartner, Deloitte, EY, Protiviti and more

Gartner's CFO surveys on AI value, then Deloitte, EY, Protiviti, RGP, BCG and others: each figure's sample, date and wording, as of November 9, 2026.

Surveys put the share of finance leaders reporting a return on AI anywhere from 14% to 90%. The low end is RGP's count of US finance executives who report meaningful ROI today. The high end is Salesforce's count of finance leaders already using AI who report a positive ROI. Many believe AI is paying off, and far fewer can show a measured return.

This page sets out what Gartner, Deloitte, EY, Protiviti, BCG and others have published since mid-2025, with each figure's sample, date and wording. Every figure links to the publisher's own page, and each page was read on November 9, 2026. Gartner's full reports are for its clients only, so the Gartner figures here come from its press releases, not the reports behind them.

What has Gartner found about how CFOs measure AI ROI?

None of Gartner's public press releases gives a survey figure on how CFOs measure AI ROI. Its finance practice's 2025–26 releases report what finance leaders say AI has delivered so far, and its analysts advise CFOs to measure realized value rather than how many pilots and tools are live. Its client reports may go further; they aren't public.

The most direct are from a March 2026 survey of 204 finance leaders. In the May 28, 2026 release, 66% of finance organizations that had adopted AI named greater efficiency and productivity as a top benefit, and 63% said AI implementation was slower than expected in 2025. The analyst's advice was that CFOs "must not mistake activity for impact." Counts of pilots, tools and use cases in production show movement, not value.

Which Gartner CFO surveys have AI figures?

Six Gartner releases since December 2025 carry a finance-leader survey figure on AI, and a seventh surveys leaders across functions. All are self-reported, and none defines ROI.

Released Who was asked, and when The figure, in Gartner's words or close to them
Dec 10, 2025 More than 200 CFOs, August 2025 Only 36% of CFOs express confidence in their ability to drive enterprise AI impact; 44% feel confident about accelerating AI use in finance
Feb 10, 2026 More than 300 CFOs and finance leaders, October 2025 Nearly 60% plan to raise finance AI investment by 10% or more in 2026; 47% put just 1% to 5% of finance technology spend on AI
Mar 23, 2026 100 CFOs, January–February 2026 Building AI talent is one of the two most challenging priorities for the next six months
May 28, 2026 204 finance leaders, March 2026 66% of AI adopters name efficiency and productivity a top benefit; 63% say implementation was slower than expected in 2025
Jul 20, 2026 The same 204 finance leaders 45% of finance AI investment leans toward productivity and 20% toward decision quality; functions with "Upend" initiatives (new value propositions, products or markets) were more than twice as likely to report high realized value
Sep 24, 2026 160 senior finance function leaders, January–April 2026 Data extraction, accounts payable and receivable automation, and report creation generally deliver returns within nine to 10 months; forecasting and insight generation take longer. Gartner's analyst calls low AI literacy the most significant barrier now
Nov 4, 2026 1,303 senior functional leaders across functions, January–April 2026 23% reported a reduction in headcount due to AI, and 17% an increase

Three things to know before quoting these:

  • The July headline and body count different things. The headline says 45% of CFOs say their AI investments lean toward productivity. The body says 45% of AI investments in finance do. Quote the body.
  • The December release has an August URL. Its page is dated December 10, 2025, and its survey ran in August 2025, but the URL starts 2025-08-12.
  • The November survey isn't a CFO survey. It asked senior leaders across functions. Its advice is aimed at CFOs: treat time AI saves as a governed resource, and connect it to outcomes such as revenue, cost avoidance or forecast accuracy, because "AI-generated capacity is a resource, not a return."

What did Gartner's CIO survey add?

Gartner's October 19, 2026 release, from its survey of 2,002 CIOs and technology executives in May to July 2026, found that only 13% of CIOs say their enterprise has realized significant or transformational value from AI so far. CFO coverage often quotes it, but the people answering were CIOs.

The same survey asked about stopping rules. Only 31% of CIOs have guidelines for when an AI pilot should not move to production, and 17% have guidelines for abandoning low-value AI initiatives already in production. CIOs with those guidelines were 2.3 and 2.6 times more likely to be what Gartner calls vanguards. That is a correlation, and Gartner defines a vanguard partly by the CEO's perception of AI value, as the CIO reports it.

What do other finance-leader surveys say about measuring AI ROI?

The share of finance leaders reporting a return on AI runs from 14% to 90% across the surveys published since mid-2025. Most of that range comes from what each survey asked.

Publisher Published Who was asked The figure
Protiviti, Global Finance Trends Aug 26, 2026 902 finance executives in North America, Europe and Asia-Pacific Only 35% say they are highly or moderately effective at measuring AI ROI; 77% of finance organizations employ AI, and 14% of those have a defined AI strategy
EY, Global DNA of the CFO Jun 3, 2026 More than 1,600 CFOs and senior finance leaders, 28 countries 49% say traditional metrics can't capture the value of technology, data, new roles or long-term investments; 50% say proving ROI upfront is a big obstacle; 68% say current metrics need redefining; 51% struggle to explain the benefits when seeking AI investment
Avalara (fielded by Censuswide) Jul 21, 2026 1,505 CFOs and senior finance leaders in the US, UK, Australia and India, June 15–22, 2026, all of whom had deployed, piloted or evaluated AI agents 92% feel moderate or significant career pressure to show AI agent ROI (50% significant); half say their AI agent initiatives have delivered only limited measurable ROI
Salesforce, CFO Priorities Sep 9, 2026 865 finance leaders in France, Germany, Japan, the UK and the US, May 4–15, 2026, double-blind Among finance leaders using AI, 90% report positive ROI
RGP Dec 12, 2025 200 US CFOs and senior finance executives at companies with $500 million or more in revenue, October–November 2025 14% report meaningful ROI today; 66% expect significant AI ROI within two years; 48% say they are ultimately responsible for AI delivering measurable value
Deloitte, Finance Trends 2026 Oct 8, 2025 1,326 CFOs or next-in-line at companies with more than $1 billion in revenue 63% have fully deployed and actively use AI; 21% report clear, measurable ROI
BCG Jun 4, 2025 More than 280 finance executives at large companies, March 2025 Median reported ROI is 10%, against the 20% many are targeting; only 45% can quantify ROI from their initiatives
Duke, Richmond Fed and Atlanta Fed, The CFO Survey Dec 17, 2025 548 respondents, November 11–December 1, 2025 AI investment is not expected to produce measurable cost savings in 2026, or to change headcount much

The Duke and Fed team's March 25, 2026 analysis, from more than 700 executives surveyed in late 2025 and early 2026, found that firms reported a 1.8% productivity gain from AI in 2025. The gain implied by their AI-attributed revenue and headcount changes was much smaller, in every major industry. The authors call it a productivity paradox.

Why do the surveys disagree so much?

They ask different questions. Salesforce counts finance leaders using AI who report a positive ROI, and its release doesn't say how large or how it was measured. Deloitte asks for clear, measurable ROI, and RGP for meaningful ROI today. BCG asks for a number and reports the median. Protiviti asks whether the organization is good at measuring at all. Read side by side, they fit together: many finance leaders believe AI is paying off, and far fewer can show a measured return.

Four other differences matter when you quote them:

  • Who's in the sample. Avalara asked only companies already working with AI agents. Deloitte asked companies with more than $1 billion in revenue, and RGP companies with $500 million or more. The Duke and Fed survey includes small firms.
  • The wording moves. RGP's own pages call its 14% "meaningful ROI" and "significant AI ROI," and its press release calls it "meaningful value." Quote the one you link to.
  • Who paid. Avalara sells AI tax compliance agents, Salesforce sells AI agents, and Gartner, Deloitte, EY, Protiviti, RGP and BCG sell advice on this problem. The Duke and Fed survey is the only one here run by academics and central banks. Roiva, which wrote this page, sells software that measures the return on AI.
  • Self-report. Every figure is what a finance leader said. None is audited, and few of these pages say how respondents calculated ROI.

What do the surveys agree on?

Three findings recur across publishers that ask differently:

  1. Adoption is well ahead of measured return. Deloitte's 63% deployed against 21% with clear ROI, and Protiviti's 77% using AI against 35% effective at measuring it.
  2. Productivity is the main benefit, and it isn't cash yet. Gartner's finance leaders name efficiency first, Gartner's November advice is to treat freed capacity as a resource rather than a return, and CFOs in the Duke and Fed survey don't expect AI to produce measurable cost savings in 2026.
  3. The method is the gap. Half of EY's CFOs say traditional metrics can't capture this value, and two-thirds want them redefined.

One more finding stands alone, so weigh it that way: few organizations have rules for stopping. Gartner's 17% of CIOs with guidelines for abandoning low-value AI in production is the only figure here on it; none of the other pages above reports one.

What should a CFO take from them?

Measure each AI initiative the way you'd measure any other investment, and decide what counts before it goes live. In practice:

  • Set the baseline and the value definition first. A baseline is the figure before the AI that later readings are compared against. Without one, a return is a guess.
  • Keep hours apart from cash. Freed capacity is hours freed, priced at what they cost. It isn't cash until the hours are redeployed or a cost actually falls, which is the point of Gartner's November advice and the Duke and Fed paradox.
  • Count full cost. Licenses, usage, people and integration, not the vendor invoice alone.
  • Write the stopping rule down. Say in advance what result would mean scaling, fixing or retiring an initiative, and record the decision when it comes.

The six-week plan covers the first board conversation, and Don't let your AI tell you the value of your AI covers what makes a figure auditable. Public companies show the same gap: in Roiva's study of 498 S&P 500 annual reports, none states an achieved, quantified return from the company's own AI use.

Where does Roiva fit?

Roiva measures some of what these surveys asked about, and not the rest.

  • Measuring AI ROI (Protiviti's 35%). Roiva tracks each AI initiative's cost, the value it delivered against a baseline, and its ROI, month by month.
  • Realized value, not deployment (Gartner, May 2026). A value figure counts only once a person approves it, or the account's auto-approve setting does, and each one says where it came from.
  • Capacity isn't a return (Gartner, November 2026; Duke and Fed). Roiva reports freed capacity beside cash value and never adds them. ROI is cash only. Hours count as cash only when a person shows they became cash and the figure is approved.
  • Stopping rules (Gartner's CIO survey). Portfolio Review brings each live initiative back on a schedule, and each decision to continue, scale, fix, pause or retire is recorded with the figures it was made on.

Roiva doesn't benchmark you against these surveys. It also doesn't measure AI literacy or skills, which Gartner's finance releases name as the main barrier, or explain how an AI agent reached a decision, which is Avalara's main concern. The methodology has the detail.

How was this roundup put together?

Roiva wrote it from each publisher's own page, read on November 9, 2026, and each figure links to that page. Roiva sells software that measures the return on AI initiatives, so a finding that CFOs struggle to measure it is good news for us. That's a reason to check this page, and the links are there so you can.

Surveys were included when a publisher's own page gives the figure, the sample and the date. Figures found only in news coverage or vendor blogs were left out, including a figure one site attributes to a Deloitte 2026 CFO survey that couldn't be traced to any Deloitte page. Where a release summarizes a report, the release is what's quoted. Corrections are welcome through the contact page.

See it against your own numbers

The free AI ROI Exposure Assessment runs Roiva's published math on your inputs — five minutes, no account, no card.

Get the next one

New field notes by email as they're published. Or subscribe to the RSS feed.