Most organisations don’t struggle with access to AI. They struggle with people actually using it.
That gap is well documented. In 2023, McKinsey reported that around one-third of organisations were using generative AI in at least one function, but only a small fraction had embedded it into everyday workflows. A year later, Deloitte found a similar pattern: high leadership interest, rising investment, and uneven day-to-day adoption.
In applied work inside large organisations, the same thing shows up. Usage data typically follows a familiar curve: a small group of enthusiastic early adopters, a long flat middle, and a sizeable group who barely touch the tools at all.
The barrier isn’t technical skill. It’s uncertainty.
People aren’t sure whether the output is good enough to trust, what’s acceptable to use in client or internal work, or whether relying on AI quietly undermines the value they bring. Faced with that ambiguity, most default to familiar ways of working.
Behavioural science helps explain why. When the personal risk of “getting it wrong” feels higher than the benefit of trying something new, avoidance is the rational response.
The organisations that see adoption rise don’t just train people on features. They reduce uncertainty. They share concrete examples from peers, set clear norms about acceptable use, and show how AI supports judgement rather than replacing it. When confidence rises, behaviour follows.
AI adoption isn’t a technology problem. It’s a behavioural one – shaped by trust, confidence, and how everyday work is designed.
