Bots are a speed layer for recurring questions. Humans are a trust layer for judgment. Winning stores write clear boundaries: what automation answers, when it stops, and who takes over.
Good for bots/AI
- Fixed prices, hours, shipping zones, catalog links
- Collecting size/color/city/qty
- Routing to the right channel
- After-hours acknowledgement with human ETA
Needs a human now
- Complaints, refunds, late shipments
- Big negotiations / wholesale
- Sensitive or angry customers
- Explicit “I want an agent” requests
Escalation in ChatLik
Pause automation on sensitive threads in Live Support/Inbox, resolve, then reactivate safely. Log why escalations happen and fix templates/knowledge — do not only blame the bot.
One-week boundary test
- Tag 30 chats: bot / mixed / human
- Find lost orders caused by inaccurate automation
- Expand bot scope only where accuracy is high
- Add an explicit “talk to human” exit
FAQ
Does AI replace humans?
No — wider FAQ coverage, still escalate complaints.
Bot lost a sale?
Fix copy/price/escalation; do not kill all automation from one case.
Related: Why automation · Live support · Inbox · AI & live support.
Operational add-on
Use the Arabic deep sections on this topic as the source of truth for SME playbooks; keep EN aligned on decisions: one primary channel, measured first-reply time, clear bot/human boundaries, and weekly template/price review before scaling spend or broadcast.
