Eric Branch — Research

Section 01

What the research shows.

The largest firms in the world are running the surveys, hosting the summits, and publishing the diagnoses. They see the gap. None of them have specified the architecture that closes it.

McKinsey
March 2025

Singla et al.

The State of AI: How organizations are rewiring to capture value

1% of executives describe their generative AI rollouts as mature.
21% of organizations using gen AI have fundamentally redesigned at least some workflows.
Of 25 attributes tested, workflow redesign had the largest effect on enterprise-level financial impact from generative AI use.

The variable that matters most is the one almost no one is changing.

McKinsey
November 2025

Singla et al.

The State of AI in 2025: Agents, Innovation, and Transformation

88% of respondents reported regular AI use in at least one business function, up from 78% a year earlier.
39% report enterprise-level impact on earnings from AI deployments.
The survey's title names the destination: agents, innovation, transformation.

Adoption is rising. Enterprise-level transformation is not following.

McKinsey
September 2025

Sukharevsky et al.

The Agentic Organization: Contours of the Next Paradigm for the AI Era

Names five pillars of the new organizational form.
Describes the destination as flat networks of hybrid agentic teams operating through continuous learning loops.
Acknowledges the insights will evolve as the technology matures.

The pillars are identified. The methodology that connects them into a working system is not.

KPMG
March 2026

AI Quarterly Pulse Survey, Q1 2026

Investment is accelerating. Barriers to ROI are rising faster.

Average projected AI investment per organization rose from $114M to $207M year-over-year.
Difficulty scaling use cases rose from 33% to 65%.
Skills gaps rose from 25% to 62%.
Difficulty quantifying long-term benefit rose from 34% to 59%.

Sample: U.S. organizations with $1B+ in annual revenue.

BCG
January 2026

Apotheker et al.

AI Radar 2026: As AI investments surge, CEOs take the lead

90% of CEOs believe AI will significantly transform what success looks like by 2028.
Sample: 640 CEOs globally.

The expectation is universal. The methodology is not.

HBR
March 2026

Lakhani, Spataro,
& Stave

The "Last Mile" Problem Slowing AI Transformation

Closed-door Harvard summit of senior leaders at large global enterprises across healthcare, banking, manufacturing, payments, and professional services.
One global payments network: 99% of employees actively use AI; finance cannot identify where the return appears as headcount shifts or cycle-time compression.
Names the Identity Problem of Tribal Knowledge as a core friction.
Names seven strategic pillars including clean-sheet process redesign, strategic knowledge capture, and agentic governance.

The maximum-effort case reports that the operating model has not changed. The article relocates the gap from design to leadership.

What the field has diagnosed but not solved.

McKinsey identifies pillars of the agentic organization but not the methodology that connects them.
Lakhani et al. name the problem and relocate the gap from design to leadership ability to imagine a different operating model.
KPMG documents AI investment nearly doubling year over year while every barrier to ROI rises sharply. The cost is in the budget. The architecture is not.
BCG documents 90% of CEOs expecting AI to redefine success by 2028 without naming what would deliver it.

Section 02

This has happened before.

The pattern is older than AI. Each technology cycle produces real value, is credited with transformation, but does not change how organizations operate. AI is the first tool capable of producing true transformation.

Era 01
1990s

Business Process Reengineering

Hammer & Champy, 1993 · Davenport, 1993

Redesigned work around what computing actually made possible. Produced real value where practiced honestly. Did not connect what those redesigned processes produced into enterprise-level intelligence. Did not change individual practice with the technology.

Era 02
2000s

Business Intelligence

Chen, Chiang, & Storey, 2012 · Davenport & Harris, 2007

Connected data across organizations into analytical capability. Produced real value. Did not redesign how the work was done. Layered on top of existing processes rather than requiring their redesign.

Era 03
Ongoing

Knowledge Management

Nonaka & Takeuchi, 1995 · Alavi, Leidner, & Mousavi, 2024

Tried to capture expertise. Faced structural limitations in storage, retrieval, and application. Could not embed knowledge structurally into operational systems. Required organizational discipline to sustain.

Each was called transformation. Each addressed one organizational level and left the others structurally untouched. AI is being deployed the same way.

Section 03

What the definition is built on.

The definition of transformation used here draws on organizational change scholarship that long predates AI. The frame, the mechanisms, and the structural argument all come from foundational academic work. AI is what makes them executable.

Burke &
Litwin
1992

Journal of
Management

A Causal Model of Organizational Performance and Change

Distinguishes transformational dynamics (mission, culture, leadership, strategy) from transactional dynamics (systems, structure, management practices).
Transformational change reaches how people think about and perform work, not merely how systems execute it.
Provides the most useful frame from the organizational change literature.

An organization that runs AI inside an unchanged operating model is producing transactional change.

Orlikowski
2000

Organization
Science

Using Technology and Constituting Structures: A Practice Lens

The same underlying technology, deployed under different architectural enactments, produces different organizational realities.
The technology is not the variable. The architecture is.

Two firms can deploy the same AI models, the same project management system, and the same communications platform, and produce completely different operational outcomes.

Mechanism
Scholarship

Senge, 1990
Nonaka & Takeuchi, 1995
Argyris & Schön, 1996

Reinforcing loops, knowledge creation, and double-loop learning.

Senge: reinforcing feedback loops as the mechanism by which learning organizations improve.
Nonaka & Takeuchi: a tacit-to-explicit-to-tacit knowledge spiral as the mechanism by which organizations create new capability.
Argyris & Schön: double-loop learning as the mechanism by which organizations correct underlying assumptions, not just outputs.

These mechanisms required organizational discipline to sustain. They ran when the organization maintained the discipline to run them. AI is the first technology capable of embedding them into operational architecture, where they run continuously as a structural property of the system itself.

Classical
Theory

Levy, 1986
Romanelli &
Tushman, 1994

Why classical transformation literature does not apply.

Levy defines second-order change as triggered by crisis, failing first-order responses, and the threat of organizational demise.
Romanelli & Tushman document transformations occurring under existential market crisis and CEO succession.
Both describe survival response to environmental disruption, not technology-driven capability development.

AI does not place organizations in decline. It does not arrive as a punctuating event. The classical frameworks describe a different phenomenon.

What the foundations establish.

Burke & Litwin's frame: transformation reaches how people think and work, not merely how systems execute.
Orlikowski's argument: the architecture is the variable, not the technology.
The mechanism scholarship: feedback loops, knowledge creation, and double-loop learning are the engines of compounding capability. AI is the first technology that can embed them structurally rather than as practices an organization must sustain.
The classical theory: crisis-driven transformation does not apply. AI requires its own definition.

The field sees the gap.