I want to draw a distinction at the outset of this issue that gets collapsed too often in boardroom conversations about AI: "AI isn't delivering value" and "we cannot prove AI is delivering value" are not the same claim, and they do not call for the same response. The data increasingly shows that most organizations are in the second category, not the first. That distinction changes what leaders should actually do about it.
Comviva's 2026 research on AI ROI measurement found that 62% of organizations report significant difficulty quantifying their AI investment at all — because AI-related expenses are scattered across cloud infrastructure, talent, data, and external vendors, with no unified accounting view pulling them together. The downstream consequence is structural: when an organization cannot accurately total what it spent, it cannot accurately calculate what it returned. The same research found total AI investment is frequently underestimated by 30 to 50 percent once hidden talent and integration costs are properly accounted for — which means a great many "positive ROI" claims circulating in boardrooms today are calculated against a denominator that is simply wrong.
The Funnel: Where ROI Visibility Disappears
The clearest way I have found to communicate this gap to a leadership team is as a funnel — not a single statistic, but the cumulative narrowing from "we are investing in AI" to "we can prove what that investment returned." Each stage of the funnel below is independently sourced, and each represents a real point of attrition between spend and proof.
SOURCES, TOP TO BOTTOM: THINK CIRCLE · IBM CEO STUDY 2026 · KPMG GLOBAL TECH REPORT 2026 · THINK CIRCLE · PWC GLOBAL CEO SURVEY 2026
Read top to bottom, the funnel tells a precise story. The overwhelming majority of organizations see real productivity gains from AI — that part of the value proposition is not in serious dispute. What collapses is the chain connecting that productivity gain to a documented, board-defensible financial return. By the bottom of the funnel, only roughly one in eight organizations can point to both higher revenue and lower costs attributable to AI. The other seven are not necessarily failing. Many of them are succeeding without the measurement infrastructure to prove it — which, to a board or an investor, is functionally indistinguishable from failing.
"AI isn't delivering value" and "we cannot prove AI is delivering value" are not the same claim. The data increasingly shows most organizations are in the second category. That distinction should change what leadership does next — measurement infrastructure, not retreat from AI investment.
— DR. VIVIAN A. ATUD · FUTURE SYSTEMS, ISSUE 20Why the Measurement Problem Is Structural, Not Incidental
It would be convenient to attribute the ROI measurement gap to laziness or lack of urgency. The data does not support that explanation. Investor pressure to demonstrate AI ROI has intensified sharply — research shows the share of organizations citing investor pressure as important or very important for proving AI returns jumped from roughly two-thirds to nine in ten in a single year. Leadership is not indifferent to this problem. It is structurally difficult to solve with the financial tooling most organizations still rely on.
Three structural factors explain most of the gap. First, cost fragmentation: AI spend is distributed across cloud infrastructure, API and licensing fees, specialized talent, data acquisition and labeling, and external vendor and consulting relationships — categories that typically report into different cost centers and different finance sub-functions, with no single owner responsible for a unified total. Second, attribution complexity: AI frequently influences outcomes through multiple touchpoints simultaneously, making it genuinely difficult to isolate what a specific AI deployment contributed versus what would have happened through other initiatives running concurrently. Third, and most underappreciated, a use-case-versus-enterprise mismatch: KPMG's research found that while the substantial majority of organizations report real value at the individual use-case level, only about a quarter can demonstrate ROI across multiple use cases — meaning most AI value today is real but localized, never aggregated into an enterprise-level number a CFO can defend to the board.
The Measurement Framework That Closes the Gap
The organizations escaping this funnel are not the ones with better AI. They are the ones who treated AI ROI measurement as its own deliberate infrastructure project, separate from the AI deployment itself. Three disciplines recur across the organizations doing this well.
Unify the cost ledger before the next deployment, not after
Every AI initiative going forward should carry a single, complete cost record spanning infrastructure, talent, data, and vendor spend — assembled at the start of the project, not reconstructed retroactively when a board asks for ROI. Organizations that wait until the board asks are the organizations whose costs come back underestimated by 30 to 50 percent, because by then the hidden categories have scattered across a dozen invoices and several quarters.
Measure at the enterprise level, not only the use-case level
A use case showing strong local ROI is necessary but not sufficient. The organizations achieving ROI across multiple use cases are the ones that built a consolidation layer — a single dashboard or reporting structure that aggregates individual use-case wins into a number the CFO can actually present. Without that layer, genuine value sits stranded inside individual teams, invisible at the level where capital allocation decisions get made.
Separate productivity metrics from financial metrics — and report both, honestly
The 79%-productivity-gains-to-12%-dual-financial-return gap in the funnel above is not a reason to stop measuring productivity. It is a reason to stop conflating it with financial ROI. Report both categories explicitly and separately. A board that understands AI is generating real productivity value that has not yet converted to documented financial return makes a fundamentally different decision than a board that believes AI has simply failed.
The leadership question this issue leaves you with: if your board asked tomorrow for your organization's total AI cost and total AI-attributable return, could finance produce both numbers within a week, audit-ready? If the honest answer is no, the gap in your organization is not whether AI is working. It is whether you have built the measurement infrastructure to know.
The 2026 AI spending environment will reward organizations that close this gap and punish, through tightening capital discipline, the organizations that do not. PwC's research is direct on the trajectory: companies proving AI's value are gaining board support for further investment, while those unable to demonstrate impact are facing increasing skepticism. That divergence compounds. Close the measurement gap this year, and next year's AI investment case gets easier to make. Defer it, and next year's case gets made under more scrutiny, with less patience, and against a board that has heard "trust the productivity gains" before.
The mandate is clear. Now it's yours to execute.
Dr. Vivian A. Atud
PhD Economist · CEO, Global Transformation Forum · African Union & APRM Consultant
AI Governance Advisor, MyClearPath AI · Bestselling Author · International Keynote Speaker
Follow Future Systems: Lead in the AI Era on LinkedIn — the intelligence your peers are reading every week.