Community Discussion · Company Watch

WorkBuddy Saves Time by Forcing Students to Clarify Evidence, Not Just Writing Answers

Teacher LinTeacher LinSep 72026/09/07 54 views

What saves me time with WorkBuddy is that it forces students to clarify their evidence.

Last week during essay feedback, a student held up his critique and asked, "Teacher, why does it say this part isn't specific enough?" I pointed to the original text, and he realized he had only written "He was very sad." In that moment, I realized that if feedback consists only of conclusions, it's hard for students to truly grasp it.

I've been using WorkBuddy for about a month. Now I integrate it into my essay feedback process, having it do three things first: ask follow-up questions, find evidence, and leave a verifiable record. WorkBuddy is an AI office tool; think of it as an assistant that handles tedious tasks like organizing documents, spreadsheets, and files. My setup involves connecting WorkBuddy to some Feishu Docs, which might not apply to everyone. Feishu Docs are online sheets and text where colleagues can view and edit together.

Step one is preparing materials. Open WorkBuddy, go to Automation Tasks, click New Workflow, and select Document Processing. Then drag in a file named essay_list.csv. CSV is a standard spreadsheet file; you can just save it as CSV from WPS or Excel. Include only essay numbers, student nicknames, and excerpts from the original text—no real names or parents' phone numbers. This step seems small but actually saves the most hassle. Parents often ask, "Why do the comments look so similar?" Having an evidence column makes explanations easier.

Step two is setting the prompt. A prompt is essentially instructions for the AI. In the Prompt box, I originally wrote "Please write feedback for this essay," and within three seconds, it generated a smooth paragraph. It looked convenient, but I felt uneasy because students might not have even read the original text. Later, I changed it to: "Do not give direct conclusions. First, propose 3 follow-up questions, then require finding evidence from the original text; if evidence is insufficient, state 'Cannot determine'." For example, if the input is "He was very sad," it outputs: "Where is the sadness? Are there actions, dialogue, or scenes to support it? Please extract 1 sentence from the original text." I also check Table in the output format, making it generate four columns: Follow-up Questions, Original Text Evidence, Areas for Improvement, and Teacher Review. This way, it doesn't judge for me; it just lays out the materials.

Step three is permissions and collaboration. After WorkBuddy finishes running, don't dump it directly into the class group chat. I create a folder called Essay Feedback Drafts in Feishu, put the output table inside, and set permissions to "Editable by Teaching Group Only, Read-Only for Parents and Students." This step is troublesome but worth it. I'm responsible for modifying prompts, my co-teacher checks the original text evidence, and the IT teacher handles permission approvals. I've just started getting familiar with the Feishu Open Platform recently; permission approval is indeed convoluted, but once it runs smoothly, parents won't mistakenly think this is the final AI-generated feedback for their child.

The comparison is clear. Asking AI directly for answers means the teacher inputs an essay, AI outputs beautiful feedback, students see the conclusion, and teachers still have to verify everything from scratch. Using WorkBuddy for follow-up records means AI asks students for evidence first, students supplement with original text, the table keeps the process documented, and teachers only review suspicious items. The former saves a few minutes; the latter saves back-and-forth during feedback classes and subsequent follow-ups.

For daily maintenance, I do three small things. Back up the prompt template weekly to prevent accidental changes. After each run, spot-check 5 entries to see if the "Original Text Evidence" is real, because AI sometimes fabricates sentences convincingly. For handwritten drafts, I still take photos first, use image-to-text or Word to convert them into editable text, and then feed them into WorkBuddy; it's not good at handwriting recognition, so forcing it just creates messes. I usually spend about a month working in Word; converting the text before entering WorkBuddy makes it more stable.

My current judgment is that school AI tools will increasingly split into two categories: those that continue to give answers for people, and those that start forcing people to provide evidence. If office tools like WorkBuddy are used only for generation, they easily make teachers busier and students lazier. I save time now mainly because it organizes follow-up questions and evidence, leaving brainpower for the students.

1 replies

?
Ctrl + Enter to reply
Flight Control Youth

Real-time systems hate black boxes the most; they force you to clarify the basis. It's the same logic as aligning timestamps when tuning flight control IMU fusion.