Management skills prove essential for working effectively with AI agents

A professor at the University of Pennsylvania ran an experiment where executive MBA students built functioning startups in just four days using tools like Claude Code, ChatGPT, and Gemini. The progress was striking: work typically spanning months was accomplished in days. However, success was not driven by advanced programming knowledge. The students had management training and expertise in their respective domains, and this distinction proved crucial. They translated clear objectives and quality criteria into effective AI instructions—a fundamental management skill. The article presents a decision-making framework for delegating tasks to AI: weighing the time required to do the work manually against processing time (writing prompts, evaluating results) and probability of success. Citing OpenAI research, the author notes that recent models like GPT-5.2 match human experts 72% of the time. The central insight is that in an era where computation is cheap and abundant, what remains scarce is knowing what to ask for. Clear communication, quality evaluation, and constructive feedback—classical elements of good management—become fundamental to working effectively with AI.