If you've ever tried to explain a formatting problem to a chatbot, you already know the frustration that comes with it. The request sounds simple: type in one layout, get another layout out. But the moment you sit down to write that prompt, the words scatter. The user isn't asking for a magic button. They're asking for a way to make their work repeatable without being rigid. That's the real challenge, and it's one worth solving.
The core issue here isn't a lack of tools. It's a mismatch between how we think about our data and how we ask for help with it. When you say "I can't use simple formulas because every balance sheet is a little different," you've already identified the real problem: your input is variable, and your output needs to follow a standard. That's not a spreadsheet problem. That's a logic problem. And the good news is that AI can handle it, but only if you give it a framework instead of a wish.
What this user needs is not a single perfect prompt. They need a prompt that describes the transformation in terms of rules, not examples. Instead of saying "convert this into the Act format," they should say something like: "Take a balance sheet where line items are listed in any order. Identify each line item by its label. Map it to the corresponding line in Schedule III. If a label doesn't match, flag it for review." That's the shift. You're not asking the AI to know your business. You're asking it to follow a mapping. And that mapping can be taught.
Tools like Power Query in Excel or even a well-structured table in Google Sheets can handle parts of this, but the real automation comes from combining those tools with an AI layer that understands context. The user doesn't need a new app. They need a better way to describe the existing one. Write the prompt as if you're training a new hire on day one: give examples of what you'll type, describe what the output should look like, and include one or two edge cases. Then test it, adjust it, and test it again. That's the workflow.
The takeaway is straightforward: stop trying to find the one perfect prompt and start building a system. Describe your input, define your output, and let the AI handle the variation. The moment you treat your prompt like a living document, the automation becomes possible. And that's the point where you stop wrestling with the tool and start getting real work done.