Managing multiple tools, APIs, and services in an enterprise environment is rarely as smooth as it looks on paper. Systems that are supposed to communicate with each other often do not, handoffs between platforms break when something upstream changes, and traditional automation approaches like Robotic Process Automation (RPA) were not built to handle unstructured inputs or dynamic decision-making. This paper presents WorkHub, an AI-native workflow automation platform developed as a final year project, designed to address these gaps without requiring a dedicated engineering team to operate it. The platform integrates a GPT-4-powered reasoning agent, a sliding-window memory module for conversational continuity, SerpAPI for live web search, conditional routing logic, and Slack-based notifications — all configurable through a visual node editor with no custom code required. We validated the system using an automated chat-response pipeline and measured end-to-end latency, routing accuracy, conversational coherence, and development effort. Results showed mean response times under 2.5 seconds, perfect routing accuracy across all 100 test executions, 94% coherence across 50 multi-turn conversations, and approximately 60% less development time compared to an equivalent Python implementation. The platform also resolves eight well-documented limitations of traditional RPA tools, making it a realistic and deployable option for teams seeking intelligent automation without deep AI engineering overhead. Keywords — intelligent workflow automation, AI agent, low-code platform, large language model, conversational memory, event-driven architecture, tool integration, conditional routing, GPT-4, robotic process automation.
Pavithra et al. (2026) studied this question.