Skip to content

johnpcutler/ai-action-plan

Repository files navigation

AI Maturity Through a Behavioral Lens

Why most AI adoption programs fail — and what to do instead.


Most organizations approach AI adoption as a technology rollout or a training problem. They buy tools, run workshops, send emails encouraging people to "embrace AI," and wonder why adoption stalls.

AI adoption is a behavior change problem. And the single most important finding from applying behavioral science to all four levels of AI maturity:

The environment is the bottleneck at every level. People are not resisting AI. They are responding rationally to environments that make AI use ambiguous, invisible, unrewarded, or unsafe. Fix the environment and the behavior follows.


Five Findings That Should Change How You Approach AI Adoption

  1. Opportunity dominates at every level. The ranking is always Opportunity > Capability > Motivation. The environment matters more than skills or attitudes.

  2. The intervention shifts as you climb. Modelling (showing examples) works at Level 1. Enablement (making things easier) works at Levels 2–3. Environmental Restructuring (changing the system itself) is required at Level 4.

  3. Persuasion ranks last everywhere. You cannot argue, incentivize, or campaign your way to AI maturity. Structural and social interventions always outperform motivational messaging.

  4. Coercion is universally counterproductive. Mandating AI adoption creates compliance theater at Level 1 and active political resistance at Level 4.

  5. Each level transition is a qualitative shift. You cannot "do more of what worked" to advance. The skills, governance, and organizational structures that succeed at Level N actively block progress to Level N+1.


The Four Levels

This analysis uses The AI Transformation Model (John Hurley, Notion) — one of the most widely adopted frameworks for understanding AI maturity (2,000+ teams):

Level Name Goal What Actually Blocks It
1 AI as a Thought Partner Help people explore ideas and improve decisions Permission vacuum — nobody said it's okay
2 AI as an Assistant Complete individual tasks faster, save employee time Neutral-to-hostile environment — gains are invisible, unrewarded
3 AI as Teammates Automate repetitive work, increase team efficiency Structural exclusion — missing platforms, unclear ownership
4 AI as the System Run critical workflows, scale organizational capacity Active immune response — political dynamics, power redistribution

Knowing where you are is only half the challenge. The harder question: why do organizations get stuck, and what actually moves them forward?


The Plan: 6–8 Quarters from Thought Partner to System

Quarter Focus Signature Move Key Milestone
Q1 Permission + first wins Senior leader publicly shares their AI use 30% weekly AI usage; guidelines published
Q2 First wins → embedding AI Integration Leads appointed per team Target workflows identified
Q3 Embedding + normalizing Cross-team showcase launches 2–3 AI-default workflows per team
Q4 Normalization → first agents First agent experiments begin Role ownership resolved; platforms available
Q5 Agent scaffolding Non-engineering team deploys first agent Multiple teams have working agents
Q6 Agent scaling + enterprise prep Cross-department pilot selected Agent governance evolving from experience
Q7 Enterprise pilot First cross-boundary AI-native process Pilot demonstrates measurable outcomes
Q8+ Enterprise scaling + identity Organizational identity work deepens Multiple enterprise processes underway

Each phase follows the same sequence: create the conditions (fix the environment) → build the capability (through practice, not training) → sustain the motivation (through identity-compatible framing and visible success).

Full plan with concrete actions, owners, success signals, and watchpoints: ai-maturity-action-plan.md


Three Rules

If you remember nothing else:

  1. Fix the environment before the people. Training and persuasion are waste when the environment is neutral or hostile. Make the behavior easy, visible, and rewarded first.

  2. Never skip levels. Each level builds foundational capability and organizational muscle that the next level depends on. Jumping from individual tool use to enterprise AI-native operations produces expensive failures.

  3. Match the intervention to the level. Lunch-and-learns don't solve agent-building capability gaps. Workflow tools don't solve political dynamics. The intervention must match the barrier.


What's In This Repo

Document What It Is Start Here If...
combined-com-b-report.md The full behavioral diagnostic — why organizations get stuck at each level, how barriers evolve, recurring patterns, intervention strategies You want to understand why each intervention is recommended
ai-maturity-action-plan.md Quarterly action plan with concrete changes, owners, success signals, and watchpoints You want to know what to do and in what order
level-1-thought-partner-com-b.md Individual COM-B diagnosis for Level 1 Your organization is working on initial AI adoption
level-2-assistant-com-b.md Individual COM-B diagnosis for Level 2 Your teams use AI but haven't embedded it into workflows
level-3-teammates-com-b.md Individual COM-B diagnosis for Level 3 You're trying to move from tool use to AI agents
level-4-system-com-b.md Individual COM-B diagnosis for Level 4 You're attempting enterprise-wide AI-native operations

How to Use This

If you're an executive:

  • Read this page and the quarter-by-quarter table above
  • Skim the "Anti-Patterns to Avoid" section in the action plan — these are the expensive mistakes

If you're leading the AI adoption effort:

  • Read the full diagnostic to understand why each intervention matters
  • Use the action plan as your operating playbook
  • Pay close attention to watchpoints and success signals

If you're a team lead:

  • Focus on the level analysis relevant to where your team is now
  • Use the success signals as your definition of done

If you're in L&D or change management:

  • The diagnostic explains why traditional approaches (awareness campaigns, training programs, executive messaging) fail for AI adoption
  • The action plan shows what to do instead

Methodology

This analysis was generated using Change Lenses and Actions — an open-source, seven-step behavior diagnosis framework built on:

  • COM-B (Capability, Opportunity, Motivation → Behavior) — the core behavioral model
  • The Behaviour Change Wheel — 9 intervention functions mapped from COM-B barriers
  • BCT Taxonomy v1 — 93 specific behaviour change techniques

The framework applies 30+ diagnostic sub-lenses across 150+ dimensional scales to each behavioral challenge. We ran the full diagnostic independently on each maturity level — examining both the adoption behavior (why organizations struggle to reach that level) and the progression blocker (why they get stuck and can't advance). The combined report synthesizes findings across all four.


What This Is Not

  • Not a technology roadmap. No specific AI tools, platforms, or vendors prescribed.
  • Not a training curriculum. Capability-building happens through scaffolded practice, not classrooms.
  • Not a one-size-fits-all timeline. Quarter estimates are directional; pace depends on context.
  • Not a mandate. The single most consistent finding is that mandating AI adoption is counterproductive.

Maturity model: The AI Transformation Model by John Hurley, Notion. Diagnostic tool: Change Lenses and Actions (John Cutler). Behavioral science: COM-B (Michie, van Stralen & West), the Behaviour Change Wheel (Michie, Atkins & West, 2014), BCT Taxonomy v1 (Michie et al., 2013).

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages