Set the AI Value Test Before the Budget Is Spent

← Back · ACT II · Chapter 8: The Buyer Lifecycle Is Progressive Governance

Eighty-four percent of finance organizations have implemented AI or plan to. Seven percent report a high or very high impact. Both figures come from a June 2025 Gartner survey of 183 CFOs, presented at the Gartner Finance Symposium in London on June 8, 2026. Adoption is close to universal. High impact is reported by fewer than one in fourteen. A separate Gartner survey of 115 CFOs in July 2026 put AI and automation second on the 2026 CFO agenda, up from fourth. So the money keeps moving toward AI while few finance leaders rate what they have already bought as high impact, and the executives signing for it hold both facts at once.

One answer now showing up in some finance functions is a dedicated value realization office. A July 2026 Forbes column by Anders Liu-Lindberg describes it as “a team whose sole job is to tie technology and transformation investments back to strategy and force an honest reckoning of whether the business case actually paid off.” The same column says the office “has explicit authority to shut down underperforming initiatives.”

The sequence matters more than the remedy. The business case was approved, the capital moved, the deployment ran, and then a new team was set up to find out whether any of it was worth the money. Every piece of governance in this arrangement arrives after the spend, so all it can do is look back.

The Reckoning Is Scheduled After the Failure

The seven percent says less about the technology than about a market that governs value after the money is gone. The claim that justified the purchase was authored by the seller, admitted without terms, and left unbound while the money did its work. Months later the enterprise asks what it got and finds the question has no referee, because no test was fixed when the claim was made. So it hires the referee now, charters the office, grants it the power to stop what is failing, and calls the arrangement discipline.

Discipline that begins after deployment can do exactly two things: measure what already happened and halt what is already failing. Both are useful, and neither of them governs value. The office can end the underperforming initiative, but the initiative entered the portfolio on an unbound claim, which means the failure was authorized at origination. Ending it later recovers the remaining budget, but not the governance that was never applied.

The office model’s answer to lateness is speed. The same Forbes column describes an office that “applies continuous governance rather than a single go/no-go decision,” watching whether a project is still delivering long after launch. That is a real improvement, but it doesn’t change what is being reviewed. A claim nobody bound doesn’t become answerable because someone looks at it sooner. Review it more often and you just find out faster that the number was never testable.

As long as claims are admitted untested, an office set up afterward can only sort out what has already been spent.

A Standard Binds the Claim at Origination

A standard does the work at the other end of the sequence. It fixes what value means, the terms under which value may be claimed, and the test every claim must pass, and it applies those terms before capital moves. The seller’s number does not become a business case until it conforms. The baseline is set at the claim, not reconstructed at the review. The conditions of realization are part of the purchase itself, not a discipline retrofitted around it.

Governed this way, the seven percent question cannot arise in its current form. Not because every investment pays, but because every claim carries its own test from the day it is made. An initiative that fails is caught by terms it agreed to at origination, not by an office that arrives two budget cycles later. Nothing has to be chartered to force an honest reckoning, because no unbound claim was admitted in the first place.

The UK has already written this order into procurement policy. On August 5, 2026, the Cabinet Office published its Social Value Model, which sets the value terms out in advance, fixes minimum weights for them in the award, and directs that what was promised be monitored through named indicators on a fixed cadence, for procurements commenced from January 1, 2027. Which parts of it bind, and how, is examined in UK Government Contracts Will Weigh Social Value and Report on It.

The Buyer Holds the Ledger

There is a reason the after-the-spend arrangement keeps failing its own audit: the claims it reviews were governed, from the start, by the party selling them. The vendor authored the number, the vendor defined the outcome, and the enterprise inherited the job of disproving it after the money was gone. An office chartered by the buyer to re-examine the seller’s story is still downstream of that story. What such an office is, as distinct from when it arrives, is the subject of a companion post, Why Offices Set Up to Track Purchase Results Keep Getting Cut.

Buyer-governed value inverts the arrangement. The buyer’s realized outcome is the only entry that settles the claim, and custody of that settlement sits in the standard by structural default. No office has to be chartered afterward and no reckoning has to be scheduled. The terms hold because they were bound when the claim was made, and they keep holding whether or not anyone is assigned to watch them.

A vendor whose product works gains from this too. A test agreed before the money moves is one it can pass on the record, instead of one an office writes after the fact.

The market has begun scheduling its honesty for the end of the project. A standard puts it at the beginning, which is the only point at which a claim can still be refused funding.

The restatement chain that carried a drifted version of this post’s opening figure into public circulation, the inversion and the re-attribution both, is now documented step by step in How One Gartner Figure Changed Meaning in Business Cases. The figures above are taken from the primary that chain failed to preserve.

Sources: Gartner press release, “Gartner Says CFOs Need Structured Finance AI Roadmaps,” June 8, 2026, reporting data from a June 2025 survey of 183 CFOs; Gartner C-level Communities, “Top 3 Priorities for CFOs in 2026,” based on 115 CFO responses to the Leadership Perspective Survey, July 2026; Anders Liu-Lindberg, “From Roadmap To ROI: How CFOs Prove Transformation Value,” Forbes, July 30, 2026; UK Cabinet Office, “PPN 026: The social value model,” August 5, 2026.

The federal government has now written the same sequence into policy: an order issued to achieve demonstrable returns puts every duty before the solicitation and assigns nobody to the proof.

Keep reading: ACT II · Chapter 8: The Buyer Lifecycle Is Progressive Governance