Why do individual AI gains not translate to business results?

McKinsey's 2026 global AI survey, conducted online May 4–June 8, 2026 with 1,719 respondents across 97 countries (weighted by each respondent's country share of global GDP, 36 percent from companies with over $1 billion in revenue), finds a split between individual and organizational payoff. "Eighty percent of respondents say AI has improved their personal productivity," the report finds, and roughly half say it helps them make better decisions and develop new skills — consistent across seniority levels. At the enterprise level, though, "only 37 percent of respondents say AI has made a positive contribution to their organization's EBIT, essentially unchanged from last year," even as the share of organizations scaling AI across the enterprise rose from 38 to 44 percent over the same period.

McKinsey's own explanation for the gap centers on a small subgroup it calls "AI high performers" — about 6 percent of respondents, who attribute at least 5 percent of EBIT to AI and describe its value as significant, a share unchanged from 2025. What separates them isn't budget size: "nearly three-quarters of high performers say they have fundamentally redesigned workflows because of AI, compared with about one-quarter of other respondents," and they are 3.3 times more likely to expect AI to transform their business within three years. High performers also build human validation into AI processes, actively manage token, compute, and storage costs, run strategic workforce planning, and report stronger senior-leadership ownership — a bundle of organizational practices, not model capability, that the report treats as the actual lever on ROI.

This sharpens a pattern already in the library. Do AI productivity gains feel larger than they actually measure? found executives' perceived AI gains outrunning measured ones; McKinsey's 80-versus-37 split is the same perception/measurement gap read through a different instrument — self-reported personal productivity against self-reported EBIT contribution, rather than perception against an independent benchmark. The organizational-practices explanation also echoes Why do ready workers struggle to transform their work?, which likewise locates the bottleneck in management and workflow design rather than individual effort or model quality. The sharper scaling gap McKinsey reports between large and small organizations — agent scaling up to 40 percent at $1B+ revenue firms versus a flat 22 percent elsewhere — parallels the firm-size concentration in Who adopts enterprise AI first and how do they use it?. McKinsey's 37 percent EBIT figure is also less stark than Why do most enterprise AI pilots fail to deliver returns?, a gap likely owing to method: a self-report survey of active AI users versus NANDA's case-based pilot analysis.

The survey measures perceptions, not audited outcomes: "positive contribution to EBIT" is a respondent's self-assessment, collected from the same pool reporting their own productivity gains, so the 80/37 contrast may partly reflect self-report bias rather than a clean individual-versus-enterprise divide. The report also cannot establish that workflow redesign or leadership ownership causes high-performer status rather than merely correlating with it — no counterfactual or longitudinal comparison is offered. What the data support is narrower: adoption and self-reported organizational payoff are decoupling at the aggregate level, and the firms narrowing that gap describe themselves as restructuring work around AI rather than layering it onto existing processes.