A distribution CFO should work out the return on one AI project from the company's own baseline, the project's full cost and a review date the CFO sets, and should test any figure from a vendor's calculator against those three.
PwC, which sells audit, tax and advisory services, published its 29th Global CEO Survey on January 19, 2026, from 4,454 CEOs in 95 countries and territories surveyed September 30 through November 10, 2025, with global figures weighted by countries' nominal GDP. Over the 12 months before the survey, more than half (56%) say their company has seen neither higher revenues nor lower costs from AI, while 12% report both. The page gives these figures for all CEOs and none for distribution. On the same page PwC, as a seller of advisory services, says its work with organisations confirms mounting evidence that isolated, tactical AI projects often deliver no measurable value, and that tangible returns come from enterprise-scale deployment consistent with business strategy. A CFO who measures one project by the steps below learns whether that project delivers measurable value.
For a project automating supplier invoice processing, the controller builds the baseline from the general ledger: the accounts payable cost centre's pay, benefits, overtime, temporary labour and outsourced processing fees for each of the 12 months before the contract is signed, so no baseline month carries the project's own hours, divided by that month's supplier invoice count from the accounts payable manager, giving a monthly cost per invoice.
The full cost is everything the project adds: the vendor's fees as the contract states them, plus the hours the CIO's team spends connecting and supporting the tool to the review date and the accounts payable team spends on testing and training, each at the loaded hourly rate the controller sets for that team. The company pays the vendor from the finance department's budget, and the CIO confirms before signing which supplier data the tool takes and where it is kept.
Before signing, the CFO sets in writing a review date no more than 12 months after go-live, a contract exit date on or soon after it, and a threshold, the lowest return the CFO will accept on the review date.
On the review date the controller subtracts each month's cost per invoice since go-live, counted without the vendor's fees or those testing and training hours, from the baseline for the same calendar month, and multiplies each difference by that month's invoice count. The sum is the saving, adjusted for any other change the controller lists with its amount and cause for CFO approval, such as a pay rise, a branch closed, an acquisition or a rise in invoice count the team could have handled without the tool. The controller adds to the saving the rise in early-payment discounts taken and the fall in late fees paid, each from its own general ledger account, each month against its counterpart before signing, adjusted for changed supplier terms and purchase volume, less interest at the company's borrowing rate on the cash paid out earlier. The return is the saving against the baseline divided by the full cost to the review date, so a return of 1 means the project saved what it cost. Connecting, testing and training mostly fall before the saving builds, so the threshold should sit below 1 for a project expected to reach 1 later. At or above the threshold the CFO keeps funding the project and, as before signing, sets in writing the next review date, the return it must reach by then, at least 1 where this threshold sat below 1, and an exit date to match. Below the threshold the CFO stops the project at the exit date.
An operating partner should ask each portfolio company's CFO for its baseline, full cost, review date, threshold and return. The CFO should test a figure from a vendor's calculator by asking for the baseline, full cost and review date behind it and replacing each with the company's own, then recomputing that return exactly as defined above. In PwC's survey, for the same 12 months, 26% of CEOs say costs have decreased due to AI, while 22% report an increase.
