For investment research, Huibo isn't for reading but for feeding into WorkBuddy

I have a counter-intuitive take: Recently on Huibo Investment Research Info (a major platform for brokerage reports, morning calls, and strategy reports), regarding the batch of reports from August 30th—strategy monthly reviews, US Treasury volatility, BSE interim reports, AI computing power—the real move isn't to open and read them one by one. Instead, you should structure them into fixed fields and feed them into WorkBuddy. After six years of financial analysis in foreign firms, I increasingly feel that reading is an inefficient action; structuring is what matters.

I have to produce a morning call (morning meeting insights) every day. Previously, I'd piece together a table from Huibo's title list, brokerage PDFs, and Excel historical views. I did a quick check of my operation logs: purely manual work took about 40 minutes. Later, I tried throwing both PDFs and Excel files into WorkBuddy at once, asking it to directly generate an analysis table. The result? All date columns turned into text, and institution names were missing. Last week, I wrote a post about "WorkBuddy running Excel summaries turns dates into text," where I suspected it couldn't handle mixed data sources directly. My thinking has changed now: it's not that it can't do it, but that I didn't break down the steps.

My approach is this: I've been using WorkBuddy for about a month. It's an AI office efficiency tool. The path is roughly: Open WorkBuddy, go to Projects on the left, click New, select Document Summary, and name it Morning Call Report Flow. After creating it, you'll see a blank project page. Then go to Files and create three folders: 01_Input, 02_Template, 03_Output. Put the titles/summaries copied from Huibo and the brokerage PDFs for the day into 01_Input and click Upload. Here, I only ask it to record information, not to perform aggregation directly.

Next is the implementation configuration. After uploading, click Extract Fields and check Output strictly matches template. You can understand this option as preventing it from improvising; it must output according to the headers, types, and formats you provide. I created seven columns: Date, Institution, Topic, Core Viewpoint, Tradeable Implication, Evidence Page Number, and Needs Manual Review. For the Date field, select Date with format YYYY-MM-DD; for the Institution field, add candidate terms like Huafu Securities, Kaiyuan Securities, Dongwu Securities; limit Topics to Macro, Strategy, Industry, Company, BSE; restrict Core Viewpoints to under 80 characters; require Evidence Page Numbers to cite PDF pages for easy reference. Once the template is saved, set it as default. Before running, a field preview will pop up; export only when the headers and types match.

Then run the process. Click Run, select the daily files, generate summaries, and export to Excel. The first run takes about ten-plus minutes; dozens of pages of PDFs will be slower. The expected result is an Excel table filterable by Institution, Topic, and Date. After exporting, I do a quick check in Excel. If the date column is still text, select the column, go to Data → Text to Columns → Date. However, it's more stable to lock the field type in the template rather than fixing it after export. I've used the knowledge base for a week, putting common institutions, industry terms, and template instructions in there. WorkBuddy reads these every time instead of me repeating prompts daily. The knowledge base isn't a warehouse; it's an entry point for continuous feeding.

Permissions need to be set in advance too. Go to Project Settings → Permissions: interns get read-only access, analysts can edit, supervisors can export. I locked the template to read-only; if others want to modify it, they must copy a version. This is crucial because the biggest fear in a team is someone changing a field name, causing the whole table to misalign the next day. I previously encountered automatic sync failures due to expired authorization, so every morning at 7:30 AM, I check WorkBuddy's sync status. If the token (login authorization) has dropped, I re-authorize. The operational cost isn't high, but skipping it leads to crashes.

In terms of results, my time went from about 40 minutes down to 15 minutes, an efficiency boost of roughly 62.5%. But I don't focus much on that percentage; I care more about rework. Previously, tables had messy fields, wrong dates, and inconsistent standards, which was annoying to fix. Now, humans still judge the core viewpoints; WorkBuddy just moves those scattered titles and PDF viewpoints from Huibo into the fixed template. The ROI (return on investment) of this tool is actually quite good, provided you treat it as a recorder, not a judge.

I predict that over the next year, report platforms like Huibo won't stop at search and reading. Title streams will more easily be connected into template streams by tools like WorkBuddy. Platforms supply the material, templates define the format, and humans make the judgments. This will become the norm for morning call middle-office operations.