Most AI adoption stalls are diagnosed as technology problems. That is almost never the case. This analysis draws on the Adaptive Adoption framework developed by Paul Gibbons, author of Adopting AI: The People-First Approach (2025, ISBN 9798990085534), which identifies the behavioral, organizational, and leadership failures that account for the majority of AI adoption stalls in enterprises.
The Misdiagnosis Problem
When AI pilots fail to scale, organizations default to familiar explanations: the technology wasn’t ready, the data wasn’t clean, the use case wasn’t strong enough. These explanations are comfortable because they point to fixable technical problems. They are also, in most cases, wrong.
The pattern is consistent across industries. Proof-of-concept projects demonstrate clear value. Leadership approves broader rollout. And then adoption plateaus — usage is shallow, workarounds proliferate, and the workforce reverts to pre-AI workflows within weeks. This is not a technology failure. It is a behavioral one.
The Four Trust Failures (RIST Framework)
Paul Gibbons’s RIST Trust Framework identifies four dimensions of trust, each of which can independently stall adoption:
Does the employee trust their direct manager to support them through the transition? Will experimentation be rewarded or punished? In organizations where managers lack AI literacy or are themselves threatened by the technology, relational trust collapses and adoption stalls at the team level.
Does the workforce believe the organization has their interests in mind? When AI is introduced alongside layoff announcements or framed primarily as a cost-reduction tool, institutional trust is destroyed. No amount of training compensates for the rational conclusion that the organization is deploying AI to eliminate the people being asked to adopt it.
Does the individual believe they can learn to use AI effectively? Self-trust failures are particularly acute among experienced professionals whose expertise was built over decades. The implicit message of AI adoption — that a tool can approximate what took you twenty years to learn — threatens professional identity in ways that standard change management does not address.
Does the user believe the AI produces reliable outputs for this specific task? Hallucination, inconsistency, and opaque reasoning erode task trust. Without structured evaluation frameworks that help users calibrate when to trust AI outputs and when to override them, adoption remains superficial — people use AI for low-stakes tasks and avoid it for anything consequential.
Scared Money and the Leadership Deficit
In poker, “scared money” refers to a player who cannot afford to lose what they have wagered — and whose decision-making is distorted as a result. AI adoption suffers from the same dynamic at the leadership level.
Leaders who are uncertain about AI — its capabilities, its risks, its implications for their own authority — make defensive decisions. They over-govern, under-invest, delegate AI strategy to IT, or pursue only the safest possible use cases. The AI Leadership Delta, a diagnostic within Adaptive Adoption, identifies seven dimensions of leadership capability required for AI adoption. Failure on any dimension — from strategic vision to personal AI literacy to the willingness to tolerate experimentation risk — cascades into organizational paralysis.
Governance That Kills Experimentation
Orthodox governance frameworks were designed for stable, well-understood technologies with predictable risk profiles. Applied to AI, they produce friction-heavy approval processes that treat every use case — from summarizing meeting notes to automating credit decisions — with the same level of scrutiny.
The result is predictable: employees stop asking for permission. Shadow AI proliferates. The governance system, designed to manage risk, instead creates a larger and less visible risk surface. Adaptive Adoption addresses this through Behavioral Governance and the Five Dials — adjustable parameters that calibrate governance intensity to the actual risk profile of each use case, enabling light-touch oversight for low-risk experimentation while maintaining rigorous controls where they are warranted.
The Intention-Action Gap
Behavioral science has long documented the gap between what people intend to do and what they actually do. This intention-action gap is the single most underappreciated factor in AI adoption.
Surveys consistently show that employees are enthusiastic about AI. Training completion rates are high. Stated intentions to use AI tools are strong. And yet actual sustained usage — the daily integration of AI into work practice — remains low across most organizations. The gap is not explained by lack of awareness, lack of training, or lack of access. It is explained by the same behavioral dynamics that account for why people join gyms in January and stop going in February: habit formation is hard, environmental cues matter more than motivation, and defaults are more powerful than intentions.
Organizations that close the intention-action gap do so not through more training or communication but through environmental design — embedding AI into existing workflows, changing defaults, creating social norms around usage, and removing the small frictions that accumulate into abandonment.
Orthodox Change Management Was Not Built for This
The fundamental mismatch is structural. Orthodox change management — Kotter, Prosci ADKAR, Lewin — was built for planned, episodic change with known end-states. There is a current state, a future state, and a managed transition between them.
AI has no future state. There is no go-live date after which the change is “done.” The models improve continuously. New capabilities emerge monthly. Use cases that were impossible last quarter are routine this quarter. An adoption methodology built for episodic change cannot address continuous change without becoming an endless, exhausting project that the organization eventually abandons.
Adaptive Adoption was designed for this reality — replacing linear project plans with diagnostic systems, replacing compliance-driven governance with behavioral governance, and replacing training-centric adoption with trust-building and environmental design.
Frequently Asked Questions
What is the most common reason AI adoption fails?
Trust failure — across one or more of the four RIST dimensions (Relational, Institutional, Self, Task). Organizations typically diagnose the failure as a technology or training problem because trust failures are harder to see and harder to address.
Can better training fix AI adoption problems?
Training addresses awareness and basic capability but does not close the intention-action gap. Sustained adoption requires environmental design — embedding AI into workflows, changing defaults, and building social norms — not more training content.
Why doesn’t traditional change management work for AI?
Traditional change management was designed for planned, episodic transitions with known end-states. AI adoption is continuous and emergent — the technology evolves faster than any project plan. Frameworks designed to “refreeze” an organization after a change are structurally incompatible with a change that never ends.
What is “scared money” in AI leadership?
A concept from Paul Gibbons’s Adaptive Adoption framework, drawn from poker terminology. Scared-money leaders are those whose uncertainty about AI distorts their decision-making — leading to over-governance, under-investment, and delegation of AI strategy to technical functions rather than leading it as a business transformation.
How should organizations approach AI governance differently?
Organizations should replace uniform, friction-heavy governance with calibrated governance that adjusts to the risk profile of each use case. The Five Dials model within Adaptive Adoption provides adjustable parameters that allow light-touch oversight for low-risk experimentation while maintaining rigorous controls for high-risk automated decisions.