78% Use AI, 80% See No Bottom-Line Impact: McKinsey's Rewiring Gap

See also (wiki): workflow-redesign · hitl-deployment-pattern · agentic-ai-governance

Executive Summary

• McKinsey’s March 2025 Global Survey on AI (n=1,491, 101 countries, July 2024 field dates) finds 78% of organizations use AI in at least one function — up from 72% six months prior — yet more than 80% report no tangible enterprise-level EBIT impact from gen AI.
• Workflow redesign is the single biggest driver of EBIT impact across all 25 organizational attributes tested. Only 21% of organizations have fundamentally redesigned any workflows when deploying gen AI.
• CEO oversight of AI governance is the attribute most correlated with bottom-line impact at large companies. Only 28% of AI-using organizations have their CEO overseeing AI governance.
• Only 1% of executives in developed markets describe their gen AI rollouts as “mature.” Fewer than one in five track well-defined KPIs for gen AI solutions — the adoption practice with the highest correlation to EBIT impact.
• Human review of gen AI outputs varies wildly: 27% review all outputs, while a similar share reviews 20% or less. Professional services firms review far more than other industries.

Temporal note: This is the March 2025 edition (July 2024 survey data). McKinsey’s subsequent November 2025 edition (Q4 2024 data, n=1,993) updates these figures: AI adoption reaches 88%, and only 6% of organizations qualify as high performers (>5% EBIT from AI). The trend line confirms: adoption accelerates, value capture does not.

Source credibility: MEDIUM. McKinsey has direct AI consulting interests. All financial impact data is self-reported. Sample skews toward large organizations (42% above $500M revenue). Methodology is consistent across McKinsey’s annual AI survey series, enabling reliable trend comparison. Directionally aligned with BCG, MIT CISR, and Deloitte independent research.

The Rewiring Gap: Adoption Without Architecture

The central finding is a widening gap between AI adoption and AI value capture. Organizations are deploying gen AI broadly — 71% use it regularly in at least one function, up from 65% six months earlier — but the deployment is running ahead of the organizational changes needed to extract value.

McKinsey tested 25 organizational attributes against self-reported EBIT impact. The top predictor: workflow redesign. Organizations that fundamentally redesign workflows when deploying gen AI see measurably higher bottom-line impact. But only 21% have done so. The rest are layering gen AI on top of existing processes — the same pattern that produced the “98% more PRs, zero delivery improvement” finding in Faros’s 2025 engineering data and the “19% slower” result in METR’s experienced-developer RCT.

The second strongest predictor at large companies: CEO oversight of AI governance. When the CEO personally owns AI governance, EBIT attribution rises. Yet only 28% of AI-using organizations report CEO-level governance ownership. In many cases, governance is split across two leaders — a structural diffusion that dilutes accountability.

The 1% Maturity Problem

In a complementary survey of developed-market executives, only 1% describe their gen AI rollouts as “mature.” This is not a measurement artifact — it aligns with the finding that fewer than one-third of organizations follow most of 12 tested adoption and scaling practices. The practice with the highest EBIT correlation — tracking well-defined KPIs for gen AI solutions — is implemented by fewer than one in five organizations.

Large companies ($500M+ revenue) are further ahead on every dimension: they are 2x more likely to have established adoption road maps, dedicated transformation teams, role-based capability training, and customer trust programs. The gap between large and mid-market organizations is widening, not closing.

Human Oversight: A Patchwork

The survey surfaces a striking range in human review of gen AI outputs. At one extreme, 27% of organizations review all gen AI content before use. At the other, a similar share reviews 20% or less. Professional services firms — where output errors carry direct client and legal liability — skew heavily toward full review. This data pairs directly with the MIT Sloan “persuasion bombing” research (Feb 2026): even when organizations mandate human review, the quality of that review degrades as LLM output volume and confidence overwhelm validators.

Where Gen AI Is Deployed

Gen AI deployment concentrates in functions where McKinsey’s prior research predicts the most value:

Text generation dominates (63% of gen AI users), followed by image generation (>33%) and code generation (>25%). Revenue increases within gen AI-deploying business units are rising. Cost reductions are now reported by a majority of respondents across most functions — a shift from early 2024 when only a minority saw cost savings.

But the enterprise-level picture is different: 80%+ see no tangible EBIT impact. The value stays trapped inside individual use cases and doesn’t propagate to the P&L.

Workforce Effects

The survey finds a pragmatic middle ground on workforce impact. A plurality (38%) predict gen AI will have little effect on overall head count in the next three years. Financial services is the only industry where respondents are more likely to predict reductions than no change. IT and product development respondents expect head count to increase.

Organizations are reskilling at scale and expect more reskilling ahead. New risk-related roles are emerging: 13% have hired AI compliance specialists, 6% have hired AI ethics specialists. AI data scientists remain the hardest role to fill — half of AI-using organizations say they need more.

McKinsey’s own analysis finds that head count reductions are one of the organizational attributes with the largest impact on bottom-line value from gen AI. This is the quiet finding in the report: organizations that reduce staff in areas where gen AI handles work see higher EBIT impact. Organizations that redeploy saved time to “entirely new activities” — the most common approach — may be diffusing the efficiency gains.

Key Data Points

What This Means for Your Organization

The pattern across every major survey — McKinsey, BCG, Deloitte, Stanford HAI — is now consistent enough to act on: organizations are adopting AI faster than they are building the organizational architecture to extract value from it. The 21% workflow-redesign rate against 78% adoption tells the story. Tools are deployed. Workflows are not redesigned. Value stays trapped.

Three actions separate organizations that capture value from those that report “no tangible EBIT impact”:

1. Redesign the workflow before (or at minimum alongside) the deployment. This is the single highest-impact organizational attribute McKinsey tested. It is also the hardest — it requires someone who understands both the current process and the AI capability well enough to redesign the handoffs, decision points, and quality gates. Most organizations skip it because it is slower than plugging in a tool.

2. Put a named executive — ideally the CEO — in charge of AI governance. Joint ownership across two leaders is the norm. It is also correlated with lower impact. One person, one throat to choke, one budget line.

3. Track KPIs from day one. The practice with the highest EBIT correlation is the one fewest organizations have implemented. If the organization cannot name the metric a gen AI deployment is supposed to move, it should not be surprised when the deployment moves nothing.

If this raised questions specific to how your organization is approaching any of these three areas, I’d welcome the conversation — brandon@brandonsneider.com.

Sources

1. McKinsey & Company / QuantumBlack. “The state of AI: How organizations are rewiring to capture value.” March 12, 2025. Survey: n=1,491, 101 countries, July 16–31, 2024. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value. Credibility: MEDIUM — self-reported data, McKinsey consulting interest, but large sample and consistent methodology across annual series.

2. McKinsey & Company / QuantumBlack. “The State of AI in 2025: Agents, Innovation, and Transformation.” November 16, 2025. Survey: n=1,993, 105 countries, Q4 2024. Supersedes this edition with updated figures (88% adoption, 6% high performers). Credibility: MEDIUM — same caveats apply.

Temporal tier: TIER 2 — Published March 2025 (Q1 2025), survey data from July 2024. Results may differ with current models and adoption patterns. The November 2025 edition provides more recent data. Cited here for unique findings on workflow redesign impact, HITL review rates, and organizational structure that the later edition does not replicate at the same granularity.

Brandon Sneider | brandon@brandonsneider.com April 2026