Abstract
Artificial intelligence is increasingly deployed within public and quasi-public service organisations facing acute resource constraints and rising demand. Yet the dominant narrative frames AI as a substitute for professional labour, with limited empirical attention to its role as an augmentative infrastructure supporting frontline advisers. This paper examines the implementation of an AI-enabled chatbot within Citizens Advice, analysing how it enhances adviser confidence, mitigates volunteer shortages, and reconfigures governance and legitimacy dynamics in a hybrid public service organisation. Rather than conceptualising AI as automation, we theorise the chatbot as a human-in-the-loop decision-support system embedded within professional service operations. Drawing on public value theory and professional legitimacy perspectives, we examine how AI augmentation affects adviser trust, role identity, and perceptions of service quality. Empirically, the study draws on organisational documentation and a large-scale Qualtrics survey of 3,400 advisers and service users, providing rare quantitative insight into frontline responses to AI in a legal advice context. Our Findings reveal three interrelated dynamics. First, AI functions as a confidence-building scaffold, particularly for non-legally trained volunteers, reducing uncertainty in routine or information-heavy cases and improving consistency of guidance. Second, trust in the chatbot is conditional and relational: confidence increases where outputs are transparent, explainable, and clearly positioned as advisory rather than determinative. Third, the technology enables resource reconfiguration, allowing scarce legally qualified advisers to concentrate on complex or high-risk cases, thereby enhancing overall service capacity. Our paper contributes in three ways. Empirically, it provides one of the largest datasets examining frontline professional responses to AI in a hybrid public legal service. Theoretically, it reframes AI adoption from substitution to augmentation, demonstrating how legitimacy is stabilised through adviser-in-the-loop governance. Practically, it identifies design principles transparency, human oversight, and for sustaining advisory capacity under volunteer scarcity, rather than displacing professional judgement.