In my MBA research with 18 UK-based SMEs, GenAI initiatives rarely failed outright. They did not collapse under pressure. They stalled. Momentum slowed not because the technology was unusable, but because interpretive and organisational conditions shifted in subtle ways. Across lower-success trajectories, five recurring patterns disrupted progress.
Five patterns that undermined progress
Leader deferral and reactive framing. Where strategic intent was unclear or communicated only in response to events, GenAI activity became fragmented. Leaders themselves expressed uncertainty about "what's the right thing?" Without clear framing, experimentation lacked organisational coherence.
Cultural fragmentation and divergent sensemaking. In several cases, distinct camps formed — enthusiasts, sceptics, avoiders. Where shared meaning failed to stabilise, uptake remained uneven and contested. The technology was the same across the organisation; the interpretation of it was not.
Absence of diffusion routines. Individual experimentation did not translate into organisational capability when there were no structured forums for sharing learning. Momentum depends on collective sensemaking, not isolated expertise. The person who figured out how to use Copilot effectively had no mechanism to make that knowledge shared.
Constraints framed as blockers rather than design parameters. Time pressure, infrastructure gaps, and data fragmentation were common across all cases. What differed was interpretation. In lower-success contexts, these constraints justified deferral. In higher-success cases, similar limitations were treated as scoping conditions — they defined the boundaries of the experiment, not the reason to delay it.
Risk and reliability concerns that eroded trust. Where validation capacity was weak, concerns about hallucination, safety, and responsibility constrained use. In some cases, this translated into outright restriction rather than staged adaptation.
The common thread
What is striking is that none of these factors were unique to struggling organisations. Structural constraints, risk concerns, and emotional responses were present in higher-success cases as well. The difference lay in how they were framed, absorbed, and collectively interpreted.
Momentum did not disappear because organisations lacked resources. It weakened when leadership sensegiving became hesitant, fragmented, or withdrawn. GenAI progress proved less dependent on early wins and more dependent on sustained interpretive orchestration under constraint.
What comes next
This concludes the nine-article series drawn from my MBA dissertation research. The central finding across all nine pieces is consistent: GenAI adoption is not primarily a technology challenge. It is a leadership and meaning-making challenge. The organisations that understood this — and acted accordingly — made the most progress.
The tenth piece in this series will mark the formal launch of Requisite Intelligence — bringing this research into practice as independent AI counsel for business leadership. If these articles have prompted questions worth exploring, I would welcome a conversation.