Consumer Adoption of AI-powered Virtual Assistants (AIVA): An Integrated Model Based on the SEM–ANN Approach

Artificial intelligence (AI) has lured consumers to orchestrate their routine activities relying on such technologies. Though AI-powered virtual assistants (AIVAs) have gained traction among service providers, these are still lagging on the demand front. This study intends to develop an ‘AIVA adoption model’ delineated under a holistic framework based on structural equation modelling and deep neural network incorporating multilayer perceptron algorithm. The sensitivity analysis designated ‘effort expectancy’ as the most dominant antecedent of AIVA adoption, followed by ‘perceived innovativeness’. While ‘perceived risk’ held high relevance, the tech users were equally concerned about the performance of AIVA in conjunction with its anthropomorphic response; however, they gave the least consideration to subjective norms. The parallel mediation analysis revealed that the adopters preferred transactional relationships with AIVA more than the communal one, while the simultaneous application of both the perspectives better generates loyal customers. The moderation analysis unveiled that the uncanny valley paradigm could not always be supportive, especially in the context of AIVA. The developed model may serve the basis to generate as well as sustain adoption and loyalty of the specified technology.