Duality of AI-based interaction systems in organisations: bridging among human and machine

Organisations can unlock AI’s full value by viewing interaction systems as socio-technical interfaces. We report two vignette experiments that jointly examine psychological and technical design choices shaping acceptance. Study 1 (N = 196) manipulated implementation strategy for AI coworkers (top-down, incremental, participatory). Participatory and transparent rollouts yielded higher perceived warmth, competence, humanness, trust, and overall acceptance than top-down approaches. Mediation analyses confirmed that warmth and competence linked humanness to trust and acceptance. Study 2 (N = 288) varied AI involvement in personnel selection (none, supportive, autonomous). Acceptance, indexed by intention to use, declined as AI autonomy increased. Perceived humanness and social presence predicted intention indirectly through performance expectancy and social influence, in line with UTAUT. Across studies, both how AI is introduced and how deeply it governs decisions shape user responses, yet social-cognitive impressions and classic acceptance factors can override structural design choices. Viewing AI systems as dual socio-technical interfaces clarifies why participatory, human-centred, and transparent integration strategies foster trust and adoption, whereas delegating high-stakes judgments to autonomous AI risks resistance. These findings extend technology-acceptance research, offer evidence-based guidance for implementation, and highlight the need to balance technical capability with human needs when embedding AI in organisational workflows in practice.