ChatGPT produced an amazingly good solution for DMCommunity’s March-2024 Challenge “AnalyzeEmployees”. Of course, this challenge deals with one of the favorite LLM application areas of Q&A. Still, it generated the correct (!) Python code based on plain English questions.

I tried to do it myself and quickly generated the same solution using Copilot. It was interesting to see what LLM does when you change only the JSON file name from “Employees.json” to “XYZ.json” (it refused to work), or when you modify the names of Employee’s attributes inside the questions (it still utilizes the old names). It is also interesting to compare this automatically generated solution with other solutions produced by traditional rules-based decision intelligence tools (Corticon, OPL, and OpenRules).