The Three-Stage Reality: What Five Major AI Maturity Frameworks Agree On

Executive Summary

• Five independent frameworks — MIT CISR (n=721 companies), BCG Build for the Future (n=1,250), McKinsey State of AI (n=1,993), Deloitte State of AI (n=3,235), and Stanford Enterprise AI Playbook (51 deployments) — converge on a single structural finding: AI adoption is not the constraint. Organizational transformation is.
• The frameworks use different labels but describe the same three-stage reality. Stage 1: AI tools deployed, minimal process change, below-average financial returns. Stage 2: workflow redesign underway, partial results, some financial impact. Stage 3: AI embedded in how work actually gets done, above-average financial performance.
• The financial stakes at each stage are quantified. MIT CISR’s academic data (n=721 companies) places Stage 1 organizations 12.6 percentage points below industry-average growth. Stage 3 organizations run 11.3 percentage points above. That is a 23.9-point spread — not a rounding difference.
• Most organizations are in Stage 1, regardless of how many AI tools they have licensed. BCG finds 60% of organizations generating minimal financial return despite active AI use. McKinsey finds only 6% hitting high-performer thresholds. Deloitte finds 37% of organizations at surface-level adoption with no meaningful process change.
• The transition from Stage 1 to Stage 3 is not a technology purchase. It is an organizational project — workflow redesign, role redefinition, outcomes measurement, and leadership commitment — that most technology vendors are not positioned to deliver and most executives have not yet scoped.

The Five Frameworks, Cross-Mapped

Each framework uses its own nomenclature. Cross-mapped against organizational behavior and financial outcome, they describe the same three-stage progression.

Source credibility ratings:

• MIT CISR: HIGH — independent academic research center, non-vendor, large cross-industry sample, financial performance measured relative to industry average
• BCG: MEDIUM — consulting advisory conflict; proprietary capability assessment methodology; cross-sectional correlation, not causal evidence; directionally corroborated by MIT CISR
• McKinsey: MEDIUM — consulting advisory conflict; self-reported EBIT attribution; large-company skew (38% above $1B revenue); directionally consistent with BCG and MIT CISR
• Deloitte: MEDIUM-HIGH — sixth-year longitudinal dataset, n=3,235, no direct financial interest in AI tool outcomes; Deloitte consulting practice benefits from findings that recommend governance investment
• Stanford Playbook: HIGH for directional findings, MEDIUM-HIGH for exact percentages — independent academic center, intentional survivorship bias (studies only successful deployments)

Stage 1: Tools Deployed, Process Unchanged

What This Stage Looks Like

Organizations in Stage 1 have AI tools. Often many. Employees are using them. Pilots have run. Some use cases have demonstrated value at the task level — faster drafting, faster search, faster coding. But the underlying workflows have not changed. AI sits on top of the existing process rather than being integrated into it.

Deloitte’s data names this precisely: 37% of surveyed organizations describe their AI use as surface-level, with minimal or no change to underlying processes. BCG puts 60% of companies in the “laggard” category — minimal financial results despite active AI use. McKinsey reports that two-thirds of the 1,993-company sample remains in pilot or experiment mode despite 88% adoption.

The Financial Result

MIT CISR quantifies what this stage produces: −12.6 percentage points of growth relative to industry average, and −9.6 pp of profit. Organizations in Stage 1 are spending on AI and running below average financial performance simultaneously. This is not evidence that AI doesn’t work — it is evidence that task-level efficiency gains, accrued inside unchanged workflows, do not aggregate into organizational financial performance.

Why Companies Stay Here

The Stanford Playbook identifies the structural cause: 77% of the hardest implementation challenges in its 51-deployment sample were organizational, not technical. Change management, data quality, process redesign, and trust-building dominate the failure mode inventory — not model selection or technical integration.

Deloitte’s readiness data makes the gap concrete: talent readiness (employee AI fluency) is at 20% among surveyed organizations. Governance readiness is at 30%. Organizations with 60% of employees holding tool access and 20% of employees with the skills to use it effectively are funding a capability they cannot extract.

Stage 2: Workflow Redesign Underway

What This Stage Looks Like

Stage 2 organizations have moved beyond surface-level adoption. Workflows are being redesigned — not all of them, but specific, high-value processes. Some financial results are appearing. The AI program has moved from IT’s domain to line-of-business ownership. Data infrastructure is improving. Governance is being formalized.

MIT CISR describes this as “Develop AI Ways of Working” — AI industrialized across the enterprise, with reusable architecture and a test-and-learn culture operating at scale. Deloitte captures 30% of surveyed organizations at this level: active process redesign, measurable productivity gains, some revenue impact.

The Financial Result

MIT CISR’s data: −3.5 pp growth and −2.2 pp profit relative to industry average. Stage 2 organizations are below average but materially better than Stage 1. More importantly, the trajectory is toward the financial break — Stage 3 is the threshold where organizations cross into above-average performance.

BCG’s companion data shows what drives the improvement: organizations redesigning workflows (not just deploying tools) capture disproportionate value because 70% of total AI value resides in people, organization, and process design. Technology and algorithms account for the remaining 30%. Stage 1 organizations compete for the 30%. Stage 2 organizations start competing for the 70%.

The Behavioral Markers

McKinsey identifies the behavioral signature of organizations on the Stage 2→3 trajectory:

• 55% of McKinsey’s “high performers” fundamentally redesigned workflows when deploying AI, versus 18% of other firms — a 3x gap that is the single most predictive behavioral difference
• C-suite ownership of AI outcomes (McKinsey: 3x higher in high performers versus others)
• Transformational ambition rather than incremental efficiency as the stated goal

Stanford’s Playbook adds a structural marker: the organizations that shipped successful deployments brought Legal, HR, Risk, and Compliance into use-case selection — before deployment was designed, not at the approval stage. This pattern avoids the resistance-as-blocker pattern that appears in 35% of deployment attempts.

Stage 3: AI Embedded in How Work Gets Done

What This Stage Looks Like

Stage 3 organizations have done the organizational work. Workflows are redesigned around AI capability, not retrofitted. Roles are redefined — Deloitte finds that 84% of organizations have not yet redesigned jobs around AI, which means Stage 3 requires crossing into the 16% that have. Governance is built into the architecture, not layered on as audit. Outcomes dashboards measure business results (cycle time, defect rate, revenue per customer), not adoption metrics.

BCG’s “future-built” companies represent 5% of the 1,250-organization sample. MIT CISR’s Stage 3 and Stage 4 combined: 64% of the 2025 sample (up from 38% in 2022). The difference in these percentages reflects real progress — the industry has been moving — but also different methodology and classification thresholds. The BCG 5% and McKinsey 6% figures are more conservative, applying stricter financial performance tests. The MIT CISR 64% reflects organizational capability assessment rather than financial outcome measurement.

The Stanford Playbook provides the most granular productivity data for Stage 3: organizations that hand AI 80%+ of the workflow (with humans handling exceptions) achieve median 71% productivity gains. Organizations keeping humans in sequential approval loops achieve 30%. The tool is the same. The workflow architecture determines the outcome.

The Financial Result

MIT CISR: +11.3 pp growth and +8.7 pp profit versus industry average for Stage 3; +17.1 pp growth and +10.4 pp profit for Stage 4. BCG documents 1.7x revenue growth and 1.6x EBIT margin for “future-built” companies versus laggards. McKinsey’s high performers (6% of 1,993 companies) achieve >5% EBIT impact attributed to AI — a threshold the non-high-performers do not cross.

The Financial Stakes Across All Five Frameworks

*The wide range reflects methodological differences. MIT CISR’s 2025 update finds 64% at Stage 3 or above — but MIT CISR’s classification is capability-based, not financial-outcome-based. BCG’s 5% and McKinsey’s 6% apply strict financial performance thresholds. The honest answer: somewhere between 5% and 34% of organizations are generating substantial, measurable financial value from AI. The majority are not.

Key Data Points

What This Means for Your Organization

The convergence across five independent frameworks is the finding. When MIT CISR (academic, n=721), BCG (consulting, n=1,250), McKinsey (consulting, n=1,993), Deloitte (consulting, n=3,235), and Stanford (academic, 51 deployments) all reach the same structural conclusion via different methodologies, the conclusion is credible: the organizations capturing AI’s financial value have redesigned how work gets done. The organizations that have not — the majority — are in Stage 1.

The practical diagnostic is simple. An organization in Stage 1 can be identified by a specific pattern: AI tools licensed and used; productivity reported at the individual task level; financial performance unchanged or deteriorating. An organization in Stage 3 can be identified by the opposite pattern: workflows redesigned, not just augmented; roles redefined, not just re-tooled; outcomes measured at the business level, not the task level.

The MIT CISR financial spread (−12.6 pp to +17.1 pp growth) is the number to put in front of a board. It converts the AI maturity question from a technology evaluation into a strategic priority with quantifiable financial consequence. The question is not “are we using AI?” The question is “are we in the 5–6% generating substantial financial return, or the 60–94% that is not?”

Stage 3 is an organizational project. The workflow redesign requires line-of-business ownership, not IT project management. The role redefinition requires HR strategy, not a training program. The outcomes measurement requires finance to define the metrics, not the AI team to count usage. These are leadership decisions, not vendor decisions. For organizations trying to map where they sit in this landscape and scope what the Stage 2→3 transition actually requires for their specific functions — that is a conversation worth having: brandon@brandonsneider.com.

Sources

1. Weill, P., Woerner, S.L., & Sebastian, I.M. (2024, December). “Building Enterprise AI Maturity.” MIT CISR Research Briefing. n=721 companies (October 2022 survey). Credibility: HIGH. Independent academic research center, financial performance measured relative to industry average controls for sector effects. URL: https://cisr.mit.edu/publication/2024_1201_EnterpriseAIMaturityModel_WeillWoernerSebastian

2. Woerner, S.L., Sebastian, I.M., Weill, P., & Káganer, E. (2025, August 21). “Grow Enterprise AI Maturity for Bottom-Line Impact.” MIT CISR Research Briefing. n=152 companies (2025 survey) + 19 executive interviews. Credibility: HIGH. URL: https://cisr.mit.edu/publication/2025_0801_EnterpriseAIMaturityUpdate_WoernerSebastianWeillKaganer

3. BCG — “The Widening AI Value Gap: Build for the Future 2025” (September 2025). n=1,250 senior executives, 9 industries, 41-capability assessment framework. Credibility: MEDIUM. Consulting advisory conflict; cross-sectional correlation, not causal evidence; directionally corroborated by MIT CISR. URL: https://www.bcg.com/publications/2025/are-you-generating-value-from-ai-the-widening-gap

4. BCG — “AI at Work 2025: Momentum Builds, but Gaps Remain” (June 2025). n=10,635 workers, 11 countries. Credibility: MEDIUM-HIGH for adoption data. URL: https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain

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

6. Deloitte — “State of AI in the Enterprise 2026” (February 2026; fieldwork August–September 2025). n=3,235 senior leaders, 24 countries. Sixth annual edition. Credibility: MEDIUM-HIGH. Largest sample; longitudinal design; Deloitte consulting practice benefits from governance/redesign recommendations. URL: https://www.deloitte.com/global/en/issues/generative-ai/state-of-ai-in-enterprise.html

7. Stanford Digital Economy Lab — “The Enterprise AI Playbook” (Pereira, Graylin, Brynjolfsson; April 2, 2026). 51 successful deployments, 41 organizations, 9 industries, 7 countries. Credibility: HIGH (directional findings); MEDIUM-HIGH (exact percentages — survivorship bias inherent). URL: https://digitaleconomy.stanford.edu/publication/enterprise-ai-playbook/

8. Fortune / Brynjolfsson — “AI Productivity Liftoff” (February 15, 2026). U.S. productivity growth 2025 data and J-Curve harvest phase context. Credibility: HIGH. URL: https://fortune.com/2026/02/15/ai-productivity-liftoff-doubling-2025-jobs-report-transition-harvest-phase-j-curve/

Brandon Sneider | brandon@brandonsneider.com April 2026