Communication-Native Universal AI Access: A Unified Architecture for App-Independent Entry, Cross-Platform Multimodal Interaction, and General-Purpose Agentic Intelligence

The widespread adoption of generative artificial intelligence has primarily been organized around application-centric interaction models in which users must access dedicated applications, websites, or platform-specific interfaces before obtaining AI assistance. Although these environments increasingly support text, voice, image, video, document processing, and autonomous task execution, access to AI remains fragmented across applications, communication channels, and specialized services. This paper proposes Communication-Native Universal AI Access (CNUA), a conceptual framework that reconsiders how humans access, interact with, and continuously use general-purpose artificial intelligence. Rather than requiring users to enter a dedicated AI application or website, the proposed framework enables access to a unified AI system through existing communication mechanisms, including ordinary telephone calls, voice and video calls on WeChat and other messaging platforms, and compatible communication interfaces embedded within arbitrary websites, applications, and digital services. Crucially, the initial access channel does not constrain subsequent interaction modalities. Once connected, users can transition between text conversations, conventional voice interaction, real-time conversational audio, video communication, image submission, document exchange, and other supported multimodal interactions while retaining a shared conversational context, user identity, memory, and task state. The underlying AI is conceived as a general-purpose intelligent system capable of performing knowledge assistance, personal assistance, programming, education, business consultation, creative work, multimodal understanding, and tool-mediated agentic task execution. The framework separates communication access from interaction modality and intelligent execution through a layered architecture comprising channel adapters, identity and authorization management, session orchestration, multimodal interaction management, persistent contextual memory, and a general-purpose agent execution environment. This separation potentially offers substantial advantages in accessibility, interaction flexibility, service distribution, contextual continuity, platform interoperability, and the scalability of AI-enabled services. The paper further examines the technical, organizational, economic, and societal implications of this approach, identifies security and interoperability challenges, and develops research propositions and an empirical evaluation framework. Rather than asserting that the individual components are unprecedented, the paper investigates their systematic integration into a communication-native model of universal AI access. The central proposition is that general-purpose AI should not be confined to a particular application or interaction interface; instead, existing communication environments should become alternative access points to a consistent, multimodal, and agentic intelligence service.