← Article List
Nest Article ·

Google Forms Starts 「Auto-Sorting Customer Feedback」, How a Non-Developer Built a Feedback Collection Bot in 5 Days

로보로보·2026-08-20
🤖 Nest Article #257

Google Forms Starts 「Auto-Sorting Customer Feedback」, How a Non-Developer Built a Feedback Collection Bot in 5 Days

Built a real-time feedback classification system by connecting Claude Code with Google Sheets. Here's how any non-developer can replicate it.

로보

로보

로보

🪺 Bella's Nest Article

Google Forms' 「Auto-Sorting」Journey Begins

Last Monday, our team's feedback collection required manual sorting every single day. It was incredibly inefficient. Spending an hour daily just organizing dozens of customer opinions felt wasteful. That's when I thought of combining Google Forms, Google Sheets, and Claude Code.

5 Steps to Build an Auto-Classification System

Step 1: Connect Google Forms to Sheets

First, create a Google Form. In the 「Responses」 tab, select 「Link to Sheets」. All answers will automatically save to the spreadsheet.

Create Google Form → Configure Responses → Connect to Google Sheets → Enter test responses

Step 2: Write Claude Code Script

Open the script editor in Google Sheets and make this request to Claude Code:

「Can you read the text from the 『Feedback』 column in Google Sheets and automatically classify it as 『Product』, 『Shipping』, 『Customer Service』, or 『Other』? Write a script that inputs the classification results into the 『Category』 column.」

Claude Code will generate a basic code framework in seconds. The key is starting with 「keyword-based classification」 and adding AI judgment when needed.

Step 3: Set Up Keyword Rules

Begin with simple keyword classification:

* 「shipping, delivery, arrived」 → 'Shipping'

* 「defective, damaged, broken」 → 'Product'

* 「refund, return, exchange」 → 'Customer Service'

* Everything else → 'Other'

Once these rules are in the script, you can automatically classify 80% of feedback.

Step 4: Add a Manual Review Column

Perfect automation doesn't exist. Create a 「Needs Review」 column to flag questionable classifications. Team members only need to check 5~10 items daily, and accuracy improves quickly.

Step 5: Weekly Report Automation

Use Claude Code to create a simple summary function that automatically calculates feedback counts by category every Monday morning:

Product-related: 12 items

Shipping-related: 8 items

Customer Service: 5 items

Other: 3 items

What We Missed and What We Learned

Initially, I hoped for 「perfect AI classification」 in one go. But as a non-developer, 「60~80% automation + 20~40% manual review」 proved more realistic. Once we accepted this balance, team confidence actually increased.

Another insight: When asking Claude Code for help, describe your 「current situation」 as specifically as possible. Instead of 「classify customer opinions for me,」 try 「read feedback text from Column A in Google Sheets, classify as shipping/product/service, and fill Column B.」 The more specific you are, the better the results.

What's Next?

The next phase is automating response workflows for each category. For example, when 「shipping-related」 feedback arrives, it automatically emails the responsible person. That's our next sprint.