When using an AI accounting agent tool, do users need to write highly detailed, engineered prompts, or can they start from a blank screen with a simple instruction?

The presenter explained that about six months ago, 'prompt engineering' put the burden on the user to craft a very precise prompt to get the desired AI output, and that's still largely true if you go straight to Claude or ChatGPT. With the tool being demonstrated (Stack), the goal is to eliminate that requirement. Instead of a perfectly engineered prompt, the user just describes a task (e.g., 'calculate and record depreciation, look for disposals, verify qualifying expenditures') and the tool, because it's built with inherent accounting knowledge, knows what to do. If it doesn't have enough information, it asks clarifying questions to make sure it executes appropriately. Users don't need to specify GL accounts or explain what a typical depreciation schedule looks like — that knowledge is built in, reducing the need for detailed, precise prompts.