Artificial intelligence is moving from supporting pharmacy work toward producing outputs that can appear to represent the pharmacist. This shift creates a boundary problem: technical capability to retrieve, classify, draft, or recommend does not by itself establish authority to perform a professional clinical act. This article proposes a Digital Pharmacist Proxy–Professional Boundary Model for distinguishing acts that may be delegated to a digital proxy, acts that require conditional human–AI co-production, and acts that must remain human-led. The model begins with the act rather than the technology and evaluates verifiability, reversibility, contextual sufficiency, relational dependence, ethical or normative load, authority and consequence, and the operational management of uncertainty and accountability. These dimensions are assembled within a proposed permission envelope specifying scope, users, population, inputs, outputs, autonomy, pharmacist oversight, fallback, monitoring, and expiry. Informational and administrative acts may be delegable when they are bounded, auditable, reversible, and low in normative consequence. Medication-related interpretation may require co-production when uncertainty, contextual incompleteness, or meaningful clinical consequence demands pharmacist review and authorization. Relational engagement, value elicitation, ethical exception handling, professional commitment, and final authority-bearing decisions remain human-led under the proposed model. Validation would require task-level technical assessment, human-factors testing, workflow evaluation, equity analysis, prospective safety study, and governance testing. The model is an original non-empirical synthesis. It does not establish professional equivalence, clinical benefit, regulatory acceptance, or deployment readiness, and its boundaries require empirical and normative validation across pharmacy settings.