Forms of applications using large language models from chatbots to AI agents
Forms of LLM-Integrated Applications from LLM-Chats to Autonomous AI Agent System
Computation and LanguageArtificial IntelligenceSoftware Engineering
Summary
Large language models, or LLMs, are used in many types of software, but it isn’t always clear what makes one kind different from another. The authors studied 22 examples from research and companies to see how these systems are built and how they work. They found seven common types, like chatbots, copilots, and AI agents, each following different patterns of control and interaction with users. This helps explain what labels like 'copilot' and 'agent' really mean in practice.
What this means in practice
- •For software developers: Design LLM-powered applications using clear architectural patterns matched to user interaction and control style described in the survey.
- •For product managers: Clarify marketing terms like copilot and agent by understanding their architectural distinctions to better position AI products for users.
A survey. It maps existing work.
Authors
Irene Weber
Abstract
Large language models (LLMs) are increasingly embedded as components in software systems, marketed under labels such as chatbot, copilot, retrieval-augmented generation, workflow, coding agent and AI agent. Whether these labels denote genuine architectural forms or serve as branding has not been assessed systematically. In the sources surveyed, labels do carry architectural content, most clearly in vendor usage: copilot denotes a router-worker architecture operating a host application under step-by-step user confirmation, while the more recent shift to the label agent coincides with AI-planned multi-step execution of which the user sees only the outcome. The coding agents of four major providers share one architecture, a reason-and-act loop delegating to subagents. This survey describes seven recurring forms---LLM chats, custom agents, retrieval-augmented generation (RAG), AI-enhanced workflows, copilots, coding agents, and, in part, agentic RAG---in a common vocabulary of agents and tools. Each is characterized along four structural dimensions (agentic RAG only partially): the architectural pattern, the control of execution and the point of user intervention, the number of agent calls per task, and tool use. An illustrative corpus of 22 systems from research publications and vendor documentation grounds the descriptions and shows where they reach their limit.