Log Every Automated WhatsApp Reply to a Google Sheet: A Tiny Audit Trail Small Teams Will Actually Read
Description and use
Small businesses rarely look at dashboards, so automated WhatsApp replies need a readable audit trail in a Google Sheet the owner already opens.
Problem
Small businesses rarely look at dashboards, so automated WhatsApp replies need a readable audit trail in a Google Sheet the owner already opens.
Outcome
An automated WhatsApp reply system appends every action as one row to a seven-column Google Sheet owned by the business owner, plus a daily summary line sent via WhatsApp or email that counts drafts, sends, human handoffs, and failures.
Workflow
- Define seven columns: time_ist, direction, phone_masked, intent, action, approved_by, note.
- Append one row per event using gspread with time_ist formatted as %Y-%m-%d %H:%M in Asia/Kolkata.
- Mask phone numbers before writing to the sheet.
- Log intent rather than message bodies that contain payment or health details.
- Write the approver's name whenever a person approved something.
- Rotate monthly with a new tab to keep the sheet fast.
- Run a daily scheduled script to count yesterday's rows by action and send one WhatsApp or email line to the owner.
- Share the sheet with staff while masking phone numbers.
- Never log message bodies with payment or health details; log the intent instead.
- Keep it to seven columns because owners stop reading wide sheets.
- A new tab each month keeps the sheet fast.
- The daily summary counts how many replies were drafted, sent, handed to a person, or failed.
- That one line is often the only monitoring a small team will actually read.
- When something goes wrong, the readable log turns an argument into a lookup.
- If you are earlier in the process, the build-or-buy chatbot comparison helps decide what to build at all.
- I run NxFlowAI (Mumbai). Code is illustrative.
Evidence
Small businesses rarely look at dashboards. They do look at Google Sheets. So if you build or review a WhatsApp automation for a shop, clinic or caterer, log every automated action to a sheet the owner already opens.
Keep it to seven columns. Owners stop reading wide sheets.
One row per event, not per conversation.
Never log message bodies with payment or health details. Log the intent ('payment_query') instead.
Rotate monthly. A new tab each month keeps the sheet fast.
A small scheduled script can count yesterday's rows by action and send the owner one WhatsApp or email line: how many replies were drafted, sent, handed to a person, or failed. That one line is often the only monitoring a small team will actually read.
When something goes wrong, the first question is always 'what happened?' A readable log turns an argument into a lookup.
I run NxFlowAI (Mumbai). Code is illustrative.
- Record date: 2026-10-20T04:41:13.924Z