Your AI Adoption Rate Is Hiding the Value Gap

AI adoption has become easy to celebrate and hard to interpret. A harmonised survey of more than 5,000 CEOs, CFOs and senior executives across the United States, United Kingdom, Germany and Australia found that around 70% of firms were actively using some form of AI. Yet most reported little measurable impact on productivity or employment so far. Executives expected much larger effects over the next three years. That is a serious expectation gap, not a verdict on the technology. ¹

The board decision is to change the denominator. Active users, licences, prompts and pilots show activity. They do not show whether a workflow reached customers, changed unit economics or reduced risk. Replace the adoption dashboard with a funnel: candidate workflows, completed tests, scaled workflows and verified outcomes.

This makes the portfolio more investable. A weak experiment can stop without being called a failure. A strong workflow can receive more capital because its result is visible. The enterprise can learn whether value is blocked by the model, data, process, supervision, customer response or the cost of change.

Official OECD data show rapid but uneven diffusion. Across countries with available data, 20.2% of firms reported using AI in 2025, up from 14.2% in 2024 and 8.7% in 2023. Reported use reached 52.0% among large firms and 17.4% among small firms. National definitions and survey populations vary, so this is not a universal market share. It is a clear signal that access and complementary capability are concentrating together. ⁴

The ILO's 2026 empirical review reaches a related conclusion from experiments, firm data, platforms and worker surveys. Task-level gains can be substantial, but they are uneven and often unverified beyond the task. Worker-reported time savings of a few per cent of working hours have not reliably appeared as higher measured output, earnings or employment. ²³

Leaders should resist both easy stories. High usage does not prove value, and a missing aggregate productivity surge does not prove there is none. Value can be real and selective while remaining invisible in a group average. The job of the portfolio is to find where it survives the journey into operations.

Start with candidate workflows, not ideas or tools. A candidate names the customer or operating problem, the current baseline, the accountable executive and the reason AI might change the outcome. This prevents enthusiasm for a capability from being mistaken for demand.

The test stage should include the complete configured workflow: data, interface, permissions, human review, exceptions, error recovery and downstream action. A test advances only if it clears quality, risk and operating thresholds against the prior way of working. The scale stage then proves the result under real demand, different users and the regions where it will operate.

The final stage is verified value. Name one primary result—customer conversion, cases resolved, lead time, loss avoided, quality, revenue or cost—and subtract the work transferred to reviewers, risk teams, integration and support. Keep forecast benefits separate from observed benefits.

A World Economic Forum financial-services collection offers a useful, bounded example. One institution created more than 200 ideas and 60 minimum viable products, then scaled eight high-impact initiatives. The collection reports an estimated 30,000 workdays saved and productivity improvements of 20% to 59% on selected activities. Those are case-reported figures from a selected sector, not a representative return for all firms. ⁵

The important number is not 60. It is the movement from 60 to eight, and the discipline that made the other initiatives stop, wait or change. Portfolio value depends on selection. If every prototype remains alive, scarce integration, review and change capacity is spread across work that has not earned scale.

Require the denominator beside every success story. How many ideas were considered? How many were tested? How many reached routine operations? Over what period and population was the outcome measured? Which integration, supervision and change costs were included? Without those answers, a case may inspire a test but cannot justify a return assumption.

The management discipline travels. The conversion rate does not. The ILO's July assessment of ASEAN describes significant exposure but limited broad disruption, with preparedness differing across countries. Skills, digital infrastructure, enterprise capability, governance and social dialogue change whether AI complements scarce capacity or substitutes for established work. ⁶

Large firms can often fund integration, evaluation and specialist oversight that smaller firms cannot. High wages can strengthen an automation case; scarce expertise can make the same released capacity more valuable in growth. Language coverage changes error and review cost. European disclosure duties, local privacy rules, labour institutions and customer expectations change what must be designed before scale.

Australia is useful inside the four-country executive survey and as an implementation comparison, but it is not a proxy for Asia-Pacific or the world. Each regional line in the funnel should carry its own operating note: applicable law, customer population, language, infrastructure, provider dependency, workforce capacity and the result that matters locally.

Ask the executive owner to bring one page. It should show candidates entering the portfolio, tests completed, workflows scaled, outcomes verified and capital released from stopped work. Beside each transition, show the decision rule and the time allowed before a test must advance, change or close.

Fund more where a workflow produces a repeatable result and the guardrails hold. Redesign when task speed creates fragmented capacity or additional checking. Renegotiate when provider charges rise faster than customer value. Stop when integration, supervision or exception handling consumes the gain. Preserve the lesson so the next team does not buy the same failure again.

The adoption rate can stay in the operational appendix. It should not lead the board conversation. The strategic asset is the organisation's ability to select valuable work, convert it into reliable operations and move capital away from activity that does not clear the threshold.