Saw 'Script Reuse Doesn't Burn Tokens,' So I Turned My WorkBuddy Monthly Report Into a Workflow

To save money on spreadsheets, the key is breaking each month's fixed actions into a repeatable process. Buyout vs. subscription is secondary. My process isn't in an Excel plugin—it's on WorkBuddy.

A few days ago I came across an article about a local Excel AI assistant called Shuyi Qingzhou.

Scripts are reusable, one-time buyout, no token burn later.

My first reaction after reading was to do the math. My second reaction was: I buy this logic, but my work isn't in a single Excel file. My monthly tables are scattered—orders in Excel, product materials in Feishu Bitable, copy in WorkBuddy. If I buy a separate tool just for Excel, that's opening another hole, and data still has to be moved around. I wrote a post at the end of July about using WorkBuddy to merge monthly Excel files. Back then I was just having it stitch 30 tables together. Now I've pushed it one step further, into a process.

First, my environment: WorkBuddy desktop version. For tables it mainly handles Excel and CSV. It doesn't touch data in images. My situation may not apply to everyone.

Step one is breaking the work apart. Open WorkBuddy, the task bar is on the left, click New Task, and you get an instruction box plus a file area. I drag the entire last month's order folder in—about 30 tables. It first lists a file manifest for me to check, confirming nothing's missing.

Step two: write only one thing per instruction. I learned this from stepping in pits. The first time, to save effort, I wrote in one sentence: "Align the headers of these tables, remove blank rows, summarize by SKU, then calculate gross margin." The result didn't match up. Later I split it: the first sentence only says "Align all tables in the folder to the same header set, remove blank rows." After it runs I check once. If it's wrong, I fix just that sentence and run again.

Step three is saving the process. After it runs successfully, there's an entry in the top right to save as a process—my version's location may not match yours, just look next to the save icon. I named it "Monthly Order Summary." When saving, a few key parameters need to be locked down: header fields fixed to six columns—SKU, color, size, cost price, sale price, inventory; date columns unified to the 2026-09-01 format; missing columns auto-filled blank without error. Once these are locked, next month I just call it directly, no need to re-explain.

Step four is drawing boundaries. Let me say a bit more here. The premise of a repeatable process is that you don't let it make judgments it can't make. My process only touches table structure—alignment, deduplication, summarization. Whether prices changed, whether colors are right, I still check myself. Before, I thought letting AI do it all in one step was most efficient. Now my thinking has changed—clear boundaries actually mean less rework. A while back I wrote a post about WorkBuddy Enterprise Skill management, and I moved that boundary-drawing thinking onto my own spreadsheet work.

On maintenance, I have a habit: at the start of each month, run the process once with an empty table first. Platforms changing field names is common. Last month Taobao changed "size" to "spec," and the headers didn't match, so the whole column got misaligned. Running an empty table once catches it in two minutes.

Let me calculate how much money you save. Subscription-based AI for spreadsheets burns tokens every run—the same work burns this month, burns again next month. Once the process is saved, the same monthly work is basically just clicking execute, with almost no extra consumption afterward.

The premise of saving tokens is that you first break the work into repeatable steps. Work that can't be broken down—no matter how cheap the tool, it's money wasted.