88% Use AI. 6% Are Winning. McKinsey's State of AI 2025.

See also (wiki): ai-maturity-models · workflow-redesign · agentic-ai-governance · board-ai-strategy

Executive Summary

• McKinsey’s State of AI 2025 (n=1,993 organizations, 105 countries, Q4 2024, published November 2025) finds 88% of organizations use AI in at least one function — up from 78% the prior year. Only 6% qualify as high performers.
• High performers are defined as organizations attributing more than 5% EBIT impact to AI use. They represent a distinct behavioral profile: 55% fundamentally redesigned workflows when deploying AI (vs. 18% of other firms), are 3.6x more likely to pursue transformational change rather than incremental efficiency, and have senior-level ownership at 3x the rate of average organizations.
• The scaling problem is universal: two-thirds of organizations remain in pilot or experiment mode. Only one-third report genuine enterprise-wide AI scaling. Only 39% can attribute any measurable EBIT impact to AI at all.
• Agentic AI is accelerating: 62% of organizations are experimenting with agents, 23% are scaling in at least one function, but under 10% have achieved enterprise-wide deployment. High performers are 3-5x more likely to be actively scaling agents than average organizations.
• AI risk is no longer theoretical: 51% of organizations experienced at least one negative AI incident in the prior 12 months. The most common: inaccuracy, compliance violations, reputational harm, and privacy failures.
• Source credibility note: McKinsey has AI consulting advisory interests. Financial impact data (EBIT attribution) is self-reported by respondents. The 6% high-performer definition (>5% EBIT from AI) is a reasonable threshold but self-reported. Cross-section correlation — high performers may succeed because of overall management quality, not AI specifically. Directionally credible, consistent with BCG and MIT CISR independent research.

The Study

Publisher: McKinsey & Company, QuantumBlack
Report title: “The State of AI in 2025: Agents, Innovation, and Transformation”
Publication date: November 16, 2025
Survey timing: Q4 2024
Sample: 1,993 organizations across 105 countries
Company profile: 38% from companies with more than $1B in annual revenue
Primary URL: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

Source credibility: MEDIUM. McKinsey is a major AI consulting firm with direct financial interest in documenting large AI value gaps. Self-reported EBIT attribution has well-documented upward bias. Sample skews toward large companies (38% above $1B revenue) — findings may not translate directly to mid-market organizations with fewer resources. That said, the sample is large, geographically diverse, and methodology consistent with McKinsey’s prior annual surveys, enabling reliable year-over-year trend comparison. BCG’s independent September 2025 study and MIT CISR’s academic maturity model reach convergent conclusions, increasing directional confidence.

Adoption Without Performance: The Core Pattern

Adoption is no longer the story. 88% of organizations use AI in at least one function — a number that will reach saturation within two reporting cycles. The relevant question is now: what does AI use actually produce?

The answer from 1,993 organizations: variable, often small, concentrated in a small minority.

• 39% report any measurable EBIT impact from AI
• Among those with EBIT impact, most report under 5% of EBIT attributable to AI
• 6% attribute 5%+ EBIT impact — the high-performer threshold
• Two-thirds remain in pilot or experiment phase despite 88% adoption

The progression from “using AI” to “scaling AI” to “AI producing meaningful financial results” is not automatic. McKinsey’s data is direct: majority adoption co-exists with majority underperformance. The bottleneck is not tool access; it is organizational deployment strategy.

What the 6% Do Differently

McKinsey’s high-performer profile is consistent across three dimensions:

Workflow Redesign

55% of high performers fundamentally redesigned workflows when deploying AI. Among other organizations, 18% did. The 3x gap is the single most predictive behavioral difference. Organizations that deploy AI into existing process designs capture marginal efficiency gains — the people and process infrastructure was built for human-only workflows, and AI sits on top of it rather than being integrated into it. Organizations that ask “how should this workflow be designed if AI is a native capability?” are capturing disproportionate value.

This finding is convergent with BCG’s 10-20-70 finding (70% of AI value from people/process, not algorithms or infrastructure) and MIT CISR’s Stage 2-to-3 transition (the financial break comes at the point organizations redesign processes rather than automate within them).

Leadership Ownership

50% of senior leaders at high-performing organizations strongly agree on AI ownership, commitment, and long-term investment. At other organizations: 16%. The 3x differential in leadership engagement precedes the financial outcome — organizations where AI is a C-suite priority operate differently than those where it is delegated to IT or a single innovation team.

Transformational vs. Incremental Ambition

High performers are 3.6x more likely to target enterprise-level transformation with AI rather than incremental efficiency improvements. They pursue growth and innovation, not just cost reduction. Organizations targeting cost reduction alone are capturing the smallest available return; the larger returns come from AI-enabled revenue uplift and capability creation. McKinsey’s function-level data makes this concrete: marketing, sales, strategy, and product development show revenue uplifts above 10%, while cost-reduction-focused functions (IT, back office) show smaller EBIT contributions.

The Scaling Problem and Agentic AI

Two-thirds of the 1,993-company sample remains in “experiment or pilot” mode despite 88% adoption. Scaling failures cluster around three structural issues:

1. Data integration: Legacy systems that cannot feed AI tools with clean, structured data prevent scaling beyond isolated deployments
2. Skills gaps: 39% cite insufficient AI skills as the primary constraint on integration
3. ROI measurement: Organizations that cannot measure AI’s contribution to business outcomes cannot make the investment case to scale

Agentic AI is entering this environment at an accelerated pace:

Agents require the organizational infrastructure that most companies haven’t built for GenAI: clean data, redesigned workflows, clear human-AI decision boundaries, and measurement systems. The 62% experimenting with agents face the same structural barrier as the 67% stuck in GenAI pilots. McKinsey does not report breakthrough agent ROI data — results are described as “10-50x efficiency improvements when implemented correctly,” a range that suggests high variance in outcomes rather than consistent value delivery.

AI Risk Is Now Operational, Not Hypothetical

51% of McKinsey’s 1,993-organization sample experienced at least one negative AI incident in the prior 12 months. The most common incident categories:

• Inaccuracy in AI-generated outputs acting as final work product
• Compliance violations (AI-assisted decisions violating regulatory requirements)
• Reputational harm (public-facing AI outputs causing external damage)
• Privacy failures (AI tools processing data beyond authorized scope)

The average organization in 2025 manages approximately 4 active AI risk types, up from 2 in 2022. Explainability — the ability to show decision logic for AI outputs — is cited as the most undermanaged risk area relative to its actual exposure level.

This matters operationally: organizations that have not built AI risk management proportional to their deployment footprint are accumulating unrecognized liability. The incident rate (51% in 12 months) is high enough that most organizations above median AI adoption will face an incident before they have governance infrastructure to respond.

Key Data Points

What This Means for Your Organization

The McKinsey data establishes a clean diagnostic. If your organization uses AI in at least one function, you are in the 88%. The more useful questions are whether you are in the 39% (any measurable EBIT impact), and whether you are tracking toward the 6% (substantial EBIT impact) or the 61% (adoption without financial results).

The behavioral profile of the 6% is specific enough to serve as an organizational audit:

• Did you redesign workflows when deploying AI, or deploy AI into existing workflows? (High performers: 55% redesigned. Others: 18%.)
• Does your CEO or COO own AI deployment outcomes, or is this delegated? (High performers: 3x higher senior ownership.)
• Are you targeting cost reduction primarily, or also growth and innovation? (High performers: transformational orientation, not incremental.)

Two-thirds of organizations are stuck in pilot mode. The scaling failure is structural — data quality, skills availability, and ROI measurement are the documented constraints, not model capability. Resolving the structural issues requires investment in infrastructure (data systems) and organization (skills, governance, measurement) rather than additional tool licenses.

The 51% incident rate is a governance signal. Most organizations at average or above AI adoption will experience a negative incident within 12 months if they haven’t already. Incident response and governance are not future problems — they are current operating requirements.

If benchmarking your organization’s position against this landscape and identifying the specific structural changes that close the gap is useful, I’d welcome the conversation: brandon@brandonsneider.com

Sources

1. McKinsey & Company — “The State of AI in 2025: Agents, Innovation, and Transformation” November 16, 2025. n=1,993 organizations, 105 countries, Q4 2024 survey. QuantumBlack authorship. URL: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai — Credibility: MEDIUM. Self-reported EBIT attribution, consulting advisory conflict, large-company skew.

2. PDF direct link: https://www.mckinsey.com/~/media/mckinsey/business functions/quantumblack/our insights/the state of ai/november 2025/the-state-of-ai-2025-agents-innovation_cmyk-v1.pdf

3. Lighthouse AI — “McKinsey State of AI 2025: Why 88% Adopt AI But Only 6% Transform Their Business” Independent analysis. URL: https://lighthouselaunch.com/blog/mckinsey-state-of-ai-2025-report-analysis — Credibility: MEDIUM (secondary analysis, no methodology addition).

4. BCG “Build for the Future 2025” (September 2025) Cross-referenced for convergent findings on workflow-redesign and 5%/6% high-performer gap. URL: https://media-publications.bcg.com/The-Widening-AI-Value-Gap-Sept-2025.pdf — Credibility: MEDIUM.

5. MIT CISR Enterprise AI Maturity Model (2025) Independent academic corroboration. n=721 companies. Cross-referenced for Stage 3 financial break and workflow-redesign requirement. See research/01-ai-native-landscape/mit-cisr-enterprise-ai-maturity-2025.md — Credibility: HIGH.

Brandon Sneider | brandon@brandonsneider.com April 2026