Generative artificial intelligence (AI) is becoming increasingly prominent in the agricultural sciences, but deciding what skills to teach in the classroom is difficult because of the rapid advancements in this emerging technology. However, despite all the advancements in recent years, a key aspect of these systems remains unchanged: leading AI models use a Python sandbox to execute tasks, from analyzing data in Microsoft Excel to building documents in Microsoft Word. This presentation discusses my experience teaching generative AI to undergraduate students in both agriculture and business contexts. To build durable skills in the domain of AI, I focus on teaching Python, not like in a traditional coding course, but to verify the actions of AI systems after the prompt has been submitted. By teaching students how to skim code written and executed by AI assistants, the resulting output becomes verifiable. I teach students the basics of standard Python packages like python-docx, python-pptx, pandas, and Matplotlib so they understand what is happening inside the AI's sandbox environment whenever uploading a Microsoft Office file to the chat. To test this comprehension, I move away from asking students to write code from scratch. Instead, assignments require students to look at the code generated by the AI to identify specific variables or figure out why a graph looked wrong. This approach shows that students do not need to be developers to use AI tools effectively. By focusing on the underlying Python logic rather than just prompt wording, educators can give students a permanent skill that will last longer than the current version of any specific AI model.