Synopsis
AI adoption has stalled in many firms, and the stall is usually diagnosed as a technology problem — faster models, better orchestration, tighter security. It almost never is. Adoption fails where fast-moving technology meets human beings and organisational systems, and the variable that governs that meeting is trust. No canonical change framework names it. The few recent models that do treat it as a happy byproduct of good leadership or better communications, and nobody in the history of the world has ever built trust by talking. Trust is a judgement, but building and repairing it is behavioural — which means there are specific actions leaders can take, and those actions belong inside any change or governance framework claiming to help with AI. This paper argues trust is not one thing. It has two failure modes, not one: undertrust produces paralysis, workarounds and avoidance, while overtrust produces uncritical acceptance of machine output. It is also multi-dimensional in a way no previous enterprise technology demanded. The RIST Framework™ decomposes it into four dimensions — Relational, Institutional, Self and Task — each with distinct failure modes requiring distinct interventions, and offers a diagnostic and operational toolkit for calibrating, building and repairing trust rather than merely proclaiming it.
Key findings
- Adoption stalls are routinely misdiagnosed as technology problems — faster models, better orchestration, improved security — when the actual failure is systemic, occurring where fast-moving technology meets human beings and organisational systems.
- No canonical change framework names trust; the recent updates that do treat it as a byproduct of good leadership or better communications — and nobody has ever built trust just by talking.
- Trust has two failure modes, not one: undertrust produces paralysis, avoidance and workarounds, while overtrust produces uncritical acceptance — so “more trust” is not the goal, calibration is.
- Leaders face four simultaneous and distinct trust challenges — in the technology, in the institutions governing it, between employees and leadership, and in workers' own capacity to keep pace — each with separate failure modes and interventions.
- Naming a variable is not managing it: culture was named in the McKinsey 7-S framework almost five decades ago and no firm became expert at changing it as a result. Trust needs to be made granular, measurable and dynamic instead.