Are LLM agents the new RPA? A comparative benchmark study of agentic automation and RPA across standardized workflow tasks

Agentic Automation with Computer Use (AACU), promise a new approach to enterprise automation. Unlike traditional Robotic Process Automation (RPA), which depends on rule-based workflows and scripting, AACU lets agents complete tasks from natural language instructions through autonomous UI interaction. This study asks whether AACU is a viable alternative to RPA. In May–June 2025 we ran controlled experiments on three standard RPA challenges (data entry, monitoring, and document extraction), comparing UiPath Studio 2023.4.0 with Anthropic’s Computer Use Agent (claude-sonnet-4-20250514) on speed, reliability, and development effort. Each process ran 10 times per technology; hypotheses were tested using Welch’s t-test and Fisher’s Exact Test. RPA was significantly faster (P2: t = −3.384, p = 0.0096; P3: t = −17.30, p < 0.001) and succeeded in 100% of runs, against 90% (P2) and 60% (P3) for AACU. A preliminary single-developer observation suggests AACU may need far less development time (P2: ∼10 vs. ∼38 minutes; P3: ∼15 vs. ∼240 minutes) and handle dynamic interfaces more flexibly, though the single-measurement design warrants caution pending a multi-developer crossover study. Current AACU implementations are not production-ready but show promise for rapid prototyping and lightweight automation. Future work should examine multi-agent orchestration, hybrid architectures, and cross-platform evaluation.