End-User Development of Oracle APEX Low-Code Applications Using Large Language Models

The natural-language understanding, code generation, and reasoning abilities of Large Language Model (LLMs) have the potential to speed up development times, especially when combined with Low-Code Development Platform (LCDPs). They could also enable end-users to make small to medium-sized changes themselves, while experienced developers can focus on the more complicated development tasks. This paper demos a prototype implementation of this concept. It enables end-users to edit Oracle Application Express (APEX) low-code applications using natural language in a chat-like user interface (UI) powered by the GPT-4 Turbo LLM. We also evaluate this prototype in a qualitative user study with APEX customers from the industry and find that they generally like both the concept and the prototype. The main problem that the study uncovered is a lack of a common vocabulary between the LLM and the users. Participants suggest to solve this by integrating support features like a glossary and an element type and name inspector into the prototype.