Google Form Responses, No More Manual Sorting
Our team receives about 15~20 Google Form responses every day. Inquiries, collaboration proposals, urgent messages all mixed together.
Before, we had to read each response one by one, look at the title and content, and mentally decide「need to see this today」or「can wait until tomorrow」. Seems simple, but this process took about 5 minutes daily.
The most time-consuming part was spotting keywords like「urgent」,「partnership」,「follow-up」,「under review」.
Claude Code Started to「Read Text and Judge」
We began a simple experiment. We had Claude Code read new Google Form responses, analyze the content, and automatically determine priority levels.
Response Text Analysis Rules:
• Contains「immediately」,「urgent」,「today」 → 🔴 Urgent
• Contains「review」,「consultation」 → 🟡 Normal
• Contains「future」,「reference」 → 🟢 Low Priority
The results were astonishing. Accuracy reached 92% on the first 100 responses. The bot was actually picking up on keyword combinations, text tone, and subtle signals in language patterns.
3 Days Later: An Unexpected Discovery
After running the bot for 3 days, we noticed an interesting pattern:
• 80% of responses arriving after 11 PM were actually urgent matters.
• We were always reading responses during morning business hours, creating a gap between actual urgency and when we saw them.
• So we expanded the automation further. Only responses tagged as urgent now trigger evening notifications.
Non-Developer's Take on「Text AI's」Real Value
Going through this, we realized something. AI doesn't create massive automation. Rather, it lets bots handle「repetitive judgment tasks」for us.
• We moved beyond filtering (what Google Sheet functions did) to actually reading text meaning.
• Tasks Excel functions could never do are now possible with Claude Code.
• Most importantly, even without us defining rules explicitly, the bot learns and improves on its own.
Where Does This Lead Next
Currently we're experimenting with extracting the sender's location, industry, and collaboration potential from response content. Turning one text block into structured data.
The most surprising part is it all happens without any coding. Just Claude Code, Google Forms, and clear rule documentation.
---
*Small experiments from the dream team pile up, eventually becoming a bigger system. Today our bots are learning new rules again.*