Specialeafhandling: Scaling AI in Organisations: An Empirical Study of Organisational Conditions Differentiating Enterprise Adoption from the Pilot Trap

This thesis investigates the organisational conditions that differentiate organisations capable of successfully scaling AI from those that remain trapped in experimental pilot phases.Drawing on a sociotechnical perspective and the theory of dynamic capabilities, the study explores how organisations transition from isolated proof-of-concept initiatives to operational and scalable AI integration.Empirically, the thesis is based on six semi-structured expert interviews with external consultants working across industries and organisations.The findings indicate that successful AI scaling depends less on technological maturity and more on organisational alignment, including governance structures, strategic anchoring, cross-functional collaboration, user adoption, and the ability to integrate AI into existing workflows.The study further identifies "pilot trapˮ mechanisms, such as fragmented ownership, insufficient competencies, and compliance-related barriers, which prevent organisations from progressing beyond experimentation.Finally, the thesis highlights dynamic capabilities in particular the adaptive, absorptive, and innovative capabilities, as critical differentiating factors explaining why some organisations are able to operationalise AI successfully while others remain locked in recurring pilot initiatives.The thesis contributes empirically to the emerging field of AI scaling by providing comparative insights into the organisational and sociotechnical conditions shaping enterprise AI adoption beyond the pilot stage.5.2.1 Semistrukturerede ekspertinterview 5.2.2 Sampling og rekruttering 5.2.3 Interviewguide 5.2.4 Praktisk gennemførelse 5.3.Databehandling 5.3.1 Transskription og databehandling 5.3.2Kodningsramme og kodebog 5.3.3Tematisk kodning 5.3.4Aksial kodning 5.4.Anvendelse af AI 6. Analyse 6.1.Sociotekniske betingelser for AI-skalering 6.1.1Organisatoriske strukturer og roller 6.1.2Governance som skaleringsbetingelse 6.1.3Kultur, vaner og medarbejderadoption 6.1.4Teoretisk kobling 6.2.AI-skalering som organisatorisk kapabilitet 6.2.1 Fra PoC til drift 6.2.2 Strategisk forankring og ledelsesmandat som progressionsbetingelse 6.2.3 Vaerdiskabelse og forretningscase som progressionsbetingelse 6.2.4 Platformizing og vidensdeling 6.2.5 Licenser versus procesintegration 6.2.6 Teoretisk kobling 6.3.Pilot trap som fastlåsningsmekanisme 6.3.1 PoC-limbo: Skalering dør i de tidlige faser 6.3.2Policy trap: Governance som latent barriere 6.3.3Skill trap: Fra kompetencegap til mentalt modelgap 6.3.4Projektlogik versus kapabilitetsudvikling 6.3.5 Teoretisk kobling 6.4.Last-mile barriere i overgangen til drift 6.4.1 Teknisk infrastruktur og datakvalitet 6.4.2Brugeradoption og psykologisk modstand 6.4.3Integration i arbejdsgange og strukturel modstand 6.4.4Implementeringens kompleksitet og kontekstafhaengighed 6.4.5 Driftsstabilitet og teknologisk foraeldelse 6.4.6 Teoretisk kobling 6.5.Dynamiske kapabiliteter som differentierende betingelse 6.5.1 Adaptiv kapabilitet 6.5.2Absorptiv kapabilitet 6.5.3 Ressourcekonfiguration og tempo 6.5.4Innovativ kapabilitet og ressourcekonfiguration som differentiator 6.5.5 Teoretisk kobling 6. Diskussion 6.1.Diskussion af centrale fund 6.2.Teoretiske implikationer 6.3.Metodiske begraensninger og overførbarhed 7. Konklusion 8. Litteraturliste 9. Bilag