When AI signalling backfires: a threshold model of perceived firm innovativeness

Purpose This study aims to interrogate the implicit monotonicity assumption in service-dominant logic by examining whether AI signalling intensity uniformly enhances perceived firm innovativeness (PFI) or exhibits threshold-dependent fragility. The study empirically validates relational coherence as a perception-level mediating state, thereby confirming its role as the integrator of AI-enabled capability inference, and specifies structural and demographic boundary conditions. Design/methodology/approach A two-study archival design tests the proposed threshold model. Study 1 exploits discrete AI capability announcements among 1,012 Asian service firms (2012–2024) using event-study difference-in-differences to identify dynamic shifts in customer text-based PFI. Study 2 estimates a dynamic panel system generalised method of moments model incorporating lagged PFI, non-linear AI intensity and relational coherence, addressing reverse causality and omitted capability shocks while computing turning points and moderated curvature. Findings AI signalling exhibits an inverted-U relationship with PFI: moderate intensity strengthens capability perception, whereas high intensity attenuates and reverses effects. Relational coherence mediates this relationship, with the indirect pathway weakening and becoming negative beyond a calculable threshold. The curvature steepens and the threshold occurs earlier in high-contact services and firms serving older customer segments. Originality/value The study advances service-dominant logic by introducing threshold-fragility into its resource-integration premise, repositioning technological operant resources as legitimacy-constrained rather than uniformly accretive. It conceptualises relational coherence as the perception-level integrator governing AI-enabled capability inference and embeds demographic and structural heterogeneity within a formally identified non-linear, dynamic analytical model structure.