WorkBuddy keeps getting case-summary categorization wrong—am I misunderstanding a step?

The problem is most likely not WorkBuddy — it's that I didn't clearly state what this step's output should look like. I'm not too sure either, so I'm here to ask.

A few days ago I came across that 2026 legal AI search tool roundup — AlphaGPT, Lexis+, Zhihe AI and the like — about extending from search all the way to case file review, case analysis, and document drafting. After reading it I didn't think about switching tools; instead I felt that the step of organizing search results after you get them is the most time-consuming. I have a labor arbitration case on hand, and similar-case search turned up a dozen-plus judgments — normally I extract the points of dispute from each one myself. After that my eyes are swimming.

I've used WorkBuddy for two months, and batch processing of documents and spreadsheets has always been smooth. This time I wanted it to categorize summaries of those dozen-plus judgments by point of dispute, then produce a summary table. I made a folder, dumped them all in, clicked that "smart organize," waited a bit, and out came a table — the categorization was basically a mess. The same point of dispute got split into three or four different names, and a few didn't make it into the table at all.

I've tried renaming files, adding numbers to files, and explicitly writing "categorize by point of dispute" in the prompt — no noticeable improvement. At one point it popped up a message, roughly "some files failed to parse, please check the format." But I opened them one by one and they're all normal docx.

So I want to ask: for this kind of categorization task, do I need to first have it summarize each judgment individually, then aggregate? Or am I dropping too many files at once? Has anyone run into this?