How One Gartner Figure Changed Meaning in Business Cases

← Back · ACT III · Chapter 21: Decision Artifacts That Govern Outcomes

A CFO who opened the Forbes CFO Network on July 30 learned that 84 percent of CFOs have not realized a return on their AI investments in finance. The figure is arresting, it is attributed to Gartner, and it is not what Gartner found.

Gartner’s June 8, 2026 release from its Finance Symposium in London reports that 84 percent of finance organizations have implemented or are planning to implement AI, and that only 7 percent report a high or very high impact. The instrument beneath it is a survey of 183 CFOs fielded in June 2025. Eighty-four percent is an adoption count. It measures how many organizations moved. It says nothing about what came back.

One step sits between that release and the Forbes summary. Gartner’s own C-level Communities property publishes the sentence “84% of CFOs say they have not yet realized a return on their AI investments in finance” and hyperlinks it directly to the release that says no such thing. Forbes read the intermediate, attributed the number to the intermediate’s separate 115-respondent survey, and placed it in front of every finance leader who opened the piece.

Three publications carried one number, it meant two different things by the end, and nobody along the chain acted in bad faith.

The Number Changed Meaning In Transit

Two mutations occurred, and they are different failures.

At the first step the meaning inverted. An adoption count became a return verdict. Those are opposite ends of an investment: one records money leaving, the other judges whether any came back.

At the second step the attribution moved. The figure detached from the 183-CFO instrument that produced it and attached to a 115-CFO instrument fielded a year later that never asked the question. A reader checking the source now lands on a survey report rather than on a survey finding.

The detail that matters most is the direction of the drift. The accurate number is worse than the restated one. Seven percent reporting high or very high impact is a harsher verdict than 84 percent reporting no return yet, because realized impact is a measurement and “not yet” is a statement of patience. The more damning figure did not survive the trip, which is what restatement usually does to a number. It keeps whatever is easiest to say in one sentence.

The Brand Carried What The Record Did Not

Gartner’s name appears at both ends of this chain. That is precisely why nobody opened the link.

Gartner’s own newsroom had drawn the distinction eleven days earlier. Its release of 28 May 2026 tells finance leaders not to mistake activity for impact, and says that counts of pilots and tools rolled out “do not prove that AI is delivering the value boards now expect.”

Buyers have been trained to treat the analyst brand as the verification step. The logo is read as the proof, and once the logo is present the underlying instrument stops being interesting. Sample size, field date, question wording and population all become invisible, which is convenient, because those four things are the entire difference between an adoption count and a return verdict.

This is the failure The Value Standard names in Chapter 4. Brand dependency replaces verification with reputation, and reputation cannot tell you what a survey asked.

The defense here was available to any reader and cost about four minutes. Follow the hyperlink. Read the sentence it lands on. The chain broke in public, in daylight, on documents anyone can open.

The Standard Binds The Buyer First

Value Standard buyers govern what the seller must produce. The Request for Value states the outcome, the evidence, and the terms the claim will be recognized against, and it puts the burden of proof where the promise originated. That binding has to happen before the money moves, and it has to bind numbers exactly like this one.

That same burden runs inward, and this is the part that gets skipped. A business case is a stack of claims. The vendor’s claims are interrogated because everyone knows the vendor is selling. The third-party figures are waved through on a footnote, and those are frequently the ones doing the persuading. A market benchmark, an analyst percentage, an industry average: these set the expectation the vendor claim is then judged against. If the benchmark is wrong, the entire comparison is wrong, and the vendor never had to lift a finger. The numbers an enterprise buys arrive with no recognition standard behind them; this chain is that general condition caught in a single, reproducible instance. Those business cases get approved on numbers like this one, and the software estate they leave behind is a receipt for every one that passed, which is the argument of the stack is the receipt.

Proof over promise is a condition the buyer’s own evidence has to satisfy before it can govern a decision, and not merely a posture taken toward sellers.

Three Questions Before A Number Enters Evidence

What instrument produced this figure. When was it fielded. What exactly was asked.

A number that survives those three questions is evidence. Anything else is a claim with a source attached to it, and attaching a source is not the same as checking one. When the answer to any of the three is a link to someone else restating the figure, the number is in circulation rather than in evidence.

Nothing about this requires a new office, a new tool, or a research budget. It requires that the buyer refuse to fund against a number it has not opened. The instrument already demands traceable proof from the seller. The buyer’s own file is the first place that demand applies. The same drift runs through figures that have nothing to do with finance: the rep-free buying statistic sales leaders quote is not quite the figure Gartner published either.

A citation is a claim about where the proof lives. Open it, or nothing has been cited.

References: Gartner, “Gartner Says CFOs Need Structured Finance AI Roadmaps,” June 8, 2026 (figures from a June 2025 survey of 183 CFOs): gartner.com. Gartner, “Gartner Says CFOs Must Stop Mistaking Finance AI Deployment for Value Creation,” 28 May 2026: gartner.com. Gartner C-level Communities, “Top 3 Priorities for CFOs in 2026,” based on 115 CFO responses, July 2026: evanta.com. Forbes CFO Network, Anders Liu-Lindberg, published July 30, 2026 as “From Roadmap To ROI: How CFOs Prove Transformation Value,” since retitled “How CFOs Turn Transformation Roadmaps Into Real ROI” at the same address: forbes.com. All three pages of the chain were re-read in full on August 15, 2026 and again on September 24, 2026, and the discrepancy is reproducible from the public pages as of the later date.

Keep reading: ACT III · Chapter 21: Decision Artifacts That Govern Outcomes