WorkBuddy messes up task extraction from morning meeting notes—is my file structure too messy?

Brothers, I just saw a roundup of 2026 AI project management tools. It mentioned task automation and workflow optimization, saying it can automatically break meeting content into to-dos. This time, I don't want to talk about repositories, but a more general question: does throwing multi-source materials together at AI actually mess up task breakdown? I've been using WorkBuddy Enterprise for a few days. I'd sporadically used some AI tools for about a month before, but having it generate a task list from a pile of materials was a first.

My idea was simple: throw in the transcript from yesterday's morning meeting recording, a few screenshots of anomalies from the project group chat, and a shift schedule Excel file. Ask it to generate who does what today and by when, organized by owner and deadline. I initially created a workspace, dragged all files in, and entered a long prompt asking it to identify owners, deadlines, and anomaly types. The result was a mess. Some tasks identified "needs rework" as "no rework needed," some missed Xiao Wang, and three groups were merged into one.

Later I tried separating them. Uploading only PDFs allowed it to extract a few items; uploading only Excel allowed field-based summarization. But mixing screenshots and text led to incomplete parsing, and exporting tables warned that some records lacked fields, preventing a unified list generation. I suspect maybe I shouldn't have thrown multi-source stuff together initially, or perhaps the instructions were too messy, so it couldn't grasp the key points.

I'm stuck on one point: Should I treat WorkBuddy as a single-source organizer, or can it really act as a task distribution assistant? Has anyone tried feeding meeting notes and group messages separately, then merging results? Would that be more stable?