JustPayMe uses AI to understand your WhatsApp messages and route you to the right action. Sometimes it gets things wrong, wrong client, wrong amount, or a tone that feels off.
When that happens, just reply to the WhatsApp conversation with one of these trigger words and we’ll record the problem:
• **flag**, generic flag • **bad**, same as flag • **wrong**, same as flag • **👎**, the thumbs-down emoji
You can add a short reason after the trigger word for context:
• *flag wrong client*, we confused two clients • *bad tone*, the message sounded rude • *wrong amount*, the figure was off • *flag made up a name*, the AI invented something
Your reply is logged into our admin triage queue. Nothing changes about the conversation you were in, you can keep going with the chase, confirm a booking, or whatever you were doing.
If you flag an AI reply but there isn’t a recent one from us, we’ll say ‘Nothing to flag’. That just means there was no AI message in the last few minutes to flag.
AI Flag Status
Flags move through a simple lifecycle: NEW (just flagged, awaiting triage) → REVIEWED (someone checked it) → FIXED (prompt updated or bug shipped) → WONTFIX (acknowledged but not actionable). Status lives on the flag, not the original audit_log row.
AI Flag Reason
You can add free text after flag/bad/wrong (e.g. 'flag wrong client'). If the reason contains a keyword we recognise, we classify it as wrong_client, wrong_amount, tone, or hallucination — otherwise it stays unclassified.
JustPayMe uses AI to understand your WhatsApp messages and route you to the right action. Sometimes it gets things wrong, wrong client, wrong amount, or a tone that feels off.
When that happens, just reply to the WhatsApp conversation with one of these trigger words and we’ll record the problem:
• **flag**, generic flag • **bad**, same as flag • **wrong**, same as flag • **👎**, the thumbs-down emoji
You can add a short reason after the trigger word for context:
• *flag wrong client*, we confused two clients • *bad tone*, the message sounded rude • *wrong amount*, the figure was off • *flag made up a name*, the AI invented something
Your reply is logged into our admin triage queue. Nothing changes about the conversation you were in, you can keep going with the chase, confirm a booking, or whatever you were doing.
If you flag an AI reply but there isn’t a recent one from us, we’ll say ‘Nothing to flag’. That just means there was no AI message in the last few minutes to flag.
AI Flag Status
Flags move through a simple lifecycle: NEW (just flagged, awaiting triage) → REVIEWED (someone checked it) → FIXED (prompt updated or bug shipped) → WONTFIX (acknowledged but not actionable). Status lives on the flag, not the original audit_log row.
AI Flag Reason
You can add free text after flag/bad/wrong (e.g. 'flag wrong client'). If the reason contains a keyword we recognise, we classify it as wrong_client, wrong_amount, tone, or hallucination — otherwise it stays unclassified.