AI Studies and ROI: Few Firms Weigh AI Value Against Cost #AIg

Principal Basil C. Puglisi reads the two newest AI value studies as one measurement gap, and he puts it plainly for clients: most companies can now say what AI costs them and that it helps, yet very few can say whether the help was worth the cost. KPMG’s Global AI Pulse Q3 2026 release, built on a survey of 2,131 senior leaders across 20 countries, reports that 61 percent of organizations review AI costs during approval and 59 percent monitor them in operation. Only 12 percent consistently assess AI value against cost across the organization, and among organizations already reporting established ROI that figure rises to 48 percent.

BearingPoint released Scaling AI for measurable impact on October 1, drawn from 1,050 C-suite executives and senior leaders in 13 countries interviewed in August 2026. Nearly three-quarters of the organizations that have implemented AI report measurable top-line or bottom-line impact, yet only 13 percent have scaled their AI initiatives fully in line with the original business case. Fewer than one-third formally assess scalability before they launch an AI initiative. Puglisi’s consulting position is that both studies describe the same missing step, because the business case names a cost and the dashboard tracks that cost, while nobody wrote down, before the work started, the outcome that would make the cost worth paying.

Factics is the method Puglisi has used since 2012 for exactly that step. Every fact leads to a tactic, every tactic leaves evidence, and the KPI is written before the action so the result can fail. His March 2026 synthesis, What 34 Reports Actually Told Us About AI, found the same structure behind the deployments that held up across that research library: a specific process, a defined metric, a baseline measurement, the deployment itself, and a post-deployment measurement. That article also set a go-live rule he still applies with clients, which is that every AI deployment needs a named human accountable for its outcomes before it launches.

KPMG separates cost control from value judgment

The full KPMG report draws a line between what it calls financial control and economic management. Watching the meter run is financial control, while judging that cost against an outcome is economic management, and only 12 percent of organizations do the second consistently, with a further 26 percent doing it for most initiatives. KPMG places the steepest drop in value realization at the same point, after organizations report value and must then realize it consistently. The headline figure looks strong by comparison, since 89 percent report measurable value from AI, and average planned AI investment for the year ahead has reached US$210 million.

The U.S. edition of the survey, which KPMG LLP released on September 24 from 314 leaders at companies with at least $1 billion in revenue, shows the approval gate tightening. Seventy-four percent now include cost reviews in AI approval processes, up from 61 percent the prior quarter, and 43 percent have usage or token budgets in place. Puglisi reads those numbers as real progress at the approval stage, where a cost review answers what the project will spend. Factics asks the second question in the same meeting: which number moves, from what baseline, by when, and who reports it.

BearingPoint finds the business case breaks at scale

BearingPoint’s maturity split shows how much the measurement habit matters. Seventy percent of the organizations it classifies as Leaders link most of their AI projects to measurable financial KPIs, compared with 34 percent of Implementers, and almost half of Leaders scale fully as planned, compared with 6 percent of Implementers. The financial results are still early. Among the 685 organizations that have implemented AI, 4 percent report revenue or service-delivery gains of at least 10 percent, while 24 percent report cost reductions of at least 10 percent.

The study also finds that 62 percent of organizations estimate AI has already created workforce overcapacity of at least 10 percent. Puglisi treats that figure as the place where an ROI claim most often overstates value. Hours saved inside a workflow reach the income statement only when a manager decides what those hours do next, and a business case that counts saved hours without naming that decision reports a return the company hasn’t banked.

BearingPoint also reports that nine in ten organizations would continue investing in AI despite limited expected ROI. That conviction can be a sound strategic bet, and Factics allows acting on it, provided the bet stays labeled as a bet in the budget so leadership can see which lines rest on evidence and which rest on belief.

What a finance lead should require before the next AI budget

KPMG writes that a control layer without a named owner is a policy rather than a management system, and its survey finds 53 percent of organizations place accountability for AI-informed decisions at C-suite level or above, including 35 percent with a named executive. Puglisi would carry that ownership one level closer to the work, to the person who runs the workflow the AI changes, because that person can see whether the outcome moved.

The consulting brief Puglisi recommends for an AI investment fits on one page. It names the workflow and its current baseline, states the outcome metric and the date it will be read, records the full run cost the project expects, and names the owner who reports the result. It also records whether the pilot was checked for scale before approval, a step BearingPoint finds fewer than one-third of organizations take. If the result beats the baseline on the review date, the project earns its next tranche of funding. If it falls short, the team re-checks the evidence first and changes the tactic second, in the order Factics prescribes, and the spend doesn’t resume until the owner decides it should.

Sources

• BearingPoint. (2026, October 1). AI delivers value, but only 13% of organizations scale it as planned [Press release]. https://www.bearingpoint.com/en-gb/about-us/news-and-media/press-releases/ai-delivers-value-but-only-13-percent-of-organizations-scale-it-as-planned/
• KPMG International. (2026, September 24). New KPMG AI Pulse Survey: As AI maturity converges, leading organizations show what AI at scale requires [Press release]. https://kpmg.com/xx/en/media/press-releases/2026/09/new-kpmg-ai-pulse-survey-as-ai-maturity-converges-leading-organizations-show-what-ai-at-scale-requires.html
• KPMG International. (2026). Global AI Pulse Q3 2026: AI at scale: Accountability, resilience and economics [Report]. https://assets.kpmg.com/content/dam/kpmgsites/xx/pdf/2026/09/global-ai-pulse.pdf
• KPMG LLP. (2026, September 24). AI’s value story sharpens as organizations gain confidence in governance, accountability and workforce adoption [Press release]. https://kpmg.com/us/en/media/news/q3-ai-pulse-2026.html

Questions readers ask

What is the KPMG Global AI Pulse Q3 2026?

The Global AI Pulse is a quarterly KPMG survey. The Q3 2026 edition draws on 2,131 senior leaders at organizations with at least US$50 million in revenue across 20 countries, with fieldwork from July 23 to August 26, 2026, and it covers AI maturity, governance, resilience, investment, and how organizations connect AI cost to value.

How many organizations weigh AI value against AI cost?

KPMG finds that 12 percent consistently assess AI value against cost across the organization, and a further 26 percent do so for most initiatives. Among organizations reporting established ROI, 48 percent assess value against cost consistently, while 61 percent of all respondents review costs at approval and 59 percent monitor them in operation.

What did BearingPoint’s Scaling AI for measurable impact study find?

BearingPoint surveyed 1,050 senior leaders in 13 countries in August 2026. Nearly three-quarters of organizations that have implemented AI report measurable top-line or bottom-line impact, yet only 13 percent have scaled AI fully in line with the original business case, and fewer than one-third assess scalability before launch.

How do AI Leaders differ from other organizations in the BearingPoint study?

Seventy percent of Leaders link most AI projects to measurable financial KPIs, compared with 34 percent of Implementers. Almost half of Leaders scale their AI initiatives fully as planned, while only 6 percent of Implementers do, which points to measurement discipline as a marker of maturity.

How does Factics apply to an AI ROI claim?

Factics requires a verified fact, a specific tactic, and a KPI defined before the action. For an AI investment, the team sets the baseline and the outcome metric before approval, reads the result on a set date, and re-checks the evidence before changing the tactic if the result falls short.

What should an AI business case include before approval?

A workable AI business case names the workflow and its baseline, the outcome metric and review date, the expected run cost, and the owner who reports the result. It also records whether the pilot was assessed for scale, so the approver can judge the expected value against the expected cost.

#AIg