Three-Layer AI Adoption Model
Coordinate company bets, team transformation, and employee experimentation
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 97%
Organize AI adoption into three connected levels. At the company level, leadership places explicit bets on areas expected to be transformed and funds pods capable of pursuing them. These bets may anticipate model improvements because capability advances can arrive faster than organizational infrastructure. At the team level, functions identify workflow-specific opportunities and combine domain knowledge with specialist support. At the employee level, the company supplies approved tools, permission, and optional education so individuals can integrate AI into their work. The layers solve different problems: strategic direction prevents scattered activity, team ownership converts technology into operational outcomes, and individual experimentation develops fluency. Progress is measured separately at each level rather than being reduced to a single usage statistic.
Origin
Extracted from Marketing Against The Grain when Kieran Flanagan divided organizational AI adoption into company, team, and employee layers.
Core principles
- 01Separate company-level bets from team and individual adoption
- 02Build for future model capability rather than today's limitations alone
- 03Give employees tools and permission to experiment
- 04Use measurable transformation goals at every layer
- 05Treat curiosity and self-directed learning as individual responsibilities
How to run it
- 1
Place company-level bets
Identify the few business areas where leadership expects AI to create transformational value and state the intended outcomes.
Pro tip Include opportunities likely to become viable as models improve.
Watch out Do not turn every possible use case into a strategic priority.
- 2
Fund outcome pods
Assign cross-functional groups with the authority, data, and resources to pursue each major bet.
Pro tip Give each pod a measurable business target rather than a technology-installation milestone.
Watch out A pod without operational ownership can become an innovation theater.
- 3
Map team workflows
Have each function identify repetitive, high-latency, high-value, or data-heavy workflows that AI could transform.
Pro tip Start with pain already visible to the people doing the work.
Watch out Central leaders should not invent detailed team use cases from a distance.
- 4
Enable individuals
Provide approved tools, safe-use rules, and accessible learning resources while encouraging self-directed experimentation.
Pro tip Share concrete internal examples that employees can adapt.
Watch out Access alone does not ensure that useful habits will form.
- 5
Measure each layer
Track strategic outcomes for major bets, workflow results for teams, and useful adoption patterns for individuals.
Pro tip Reward demonstrated improvements rather than raw prompt or login counts.
Watch out A high volume of AI use may coexist with little business value.
- 6
Spread proven practices
Turn successful experiments into reusable projects, workflows, and training that other teams can adopt.
Pro tip Preserve the domain context that made the original use case effective.
Watch out Avoid forcing a workflow onto teams whose needs materially differ.
In the wild
Leadership makes localization a company-level AI bet and assigns a specialist pod to build the workflow. Regional marketing teams contribute language rules, review criteria, and edge cases. Individual marketers receive approved AI tools for lower-risk experimentation in drafting and research. The company tracks localization speed and quality separately from general employee adoption.
→ Strategic investment, domain ownership, and individual learning reinforce one another without collapsing into a single centralized program.
Common mistakes
Measuring only tool adoption
Seat activation and prompt counts do not show whether strategic bets or team workflows are improving.
Relying on only one layer
Top-down programs lack local context, while unsupported grassroots experiments struggle to scale or meet security requirements.
Is it for you?
Best for
Organizations that need a coherent operating model for AI experimentation and transformation across multiple functions.
Not ideal for
Very small teams where company, functional, and individual ownership are effectively the same layer.
From the transcript
“Um, I kind of divide it up into three parts in terms of AI adoption within the company.”
“Then the second thing is team enablement, and then the third thing is employee in general enablement.”
“I think on the employee side of things, the job of the team and the company is to provide the tools and the kind of…”
From the episode
How a $1B+ Crypto Company Really Uses AI in Marketing