The Model
Put simply: people only go along with change if they can, want to, and are allowed to. Leadership acts on this through three levers: giving direction, enabling involvement, and leading by example with credibility. Wherever a lever crosses a condition, a concrete leadership principle emerges - that's what makes up the matrix below.
Three leadership levers, crossed with three behavioural conditions, framed by four phases. The matrix is the map. It holds true throughout. The phases are the route. The focus moves.
The model is an adaptation of established approaches: COM-B as a behavioural framework. Phase logic modelled on the classic three-part structure Prepare / Execute / Anchor, as well as on ADKAR.
Direction
Create meaning, make the goal tangible.
Involvement
Make people part of the solution.
Credibility & Ownership
Lead by example instead of decreeing, don't delegate it away.
Capability
the ability to do it.
Motivation
the motivation to do it.
Opportunity
the opportunity to do it.
Interactive matrix
Optionally select a phase to see which fields are in focus there. Click a field to see details.
Governance is part of the model, not an appendix
Two of the nine fields concern rules: defining roles and guardrails early, and later making governance lastingly effective through impact measurement. This double placement is deliberate. Governance considered only at the end merely legitimises the isolated solutions that have emerged in the meantime.
The normative framework for this has existed since 2023 in ISO/IEC 42001, the first international standard for AI management systems, adopted in Switzerland as SN EN ISO/IEC 42001:2026. Two of its requirements are central to this guide: top management is not addressed as a delegation instance but as an accountable party subject to evidence requirements. And involving those affected is not a recommendation but a requirement.
So the standard says the same thing as the survey, just in different language: responsibility cannot be handed off to IT, and involvement is not a courtesy.
What this guide does not do
Four limitations belong here.
The guide starts once the decision to introduce AI has already been made. Whether and for what purpose a company should introduce AI at all is not something it answers - that's a technology and strategy decision, not a leadership question in the narrower sense. Anyone still unsure here should make that decision first, before the change begins.
A survey from a single company and a public benchmark survey allow for a reasoned assessment, but not a generalisation. The survey values are also based on self-assessment; socially desirable answering behaviour cannot be ruled out.
The model is an adaptation, not an invention. The behavioural logic comes from the COM-B model, the phase logic from the classic three-part structure of preparing, executing, and anchoring. What's original is the lever structure and its crossing with the behavioural conditions, not the framework itself.
The assessment is a self-assessment, not a measuring instrument. The thresholds are set, not derived from comparative data. They're useful for directing attention, not for proving a state of affairs.
And one limit remains that can't be planned away: the guide assumes that leadership shows the same commitment for months on end as on day one. Whether it does isn't decided by the quality of a model, but by its behaviour.
Questions about the model
Answers are generated by a language model and may contain errors.