// Retail chain · Singapore
Customer Conversation SLA Monitor
Every customer message on WhatsApp and email is ingested, classified by AI and tracked against an SLA, so no lead is left waiting.
- Client
- Retail chain · Singapore
- Built at
- WithMi Automation
- Role
- Solution design & development
- n8n
- 8 workflows · 118 nodes
The problem
Sales staff handled customers across several WhatsApp numbers and a shared inbox. Management had no visibility into which conversations were still waiting for a reply, so messages slipped through and deals were lost to slow responses.
The solution
A monitoring layer of eight workflows. Inbound WhatsApp and email messages are deduplicated, matched to a staff identity, scrubbed of personal data and classified by an AI agent into business categories. Outbound monitors detect when staff reply, a scheduled SLA engine flags conversations that breach their response window, and a daily morning report summarizes the previous day for management on WhatsApp and email.
Architecture
Interactive view of the actual workflows. Drag to pan, pinch or use + / − to zoom, click a node for details, or press Run to simulate an execution.
19 nodes · Click a node to inspect it
Engineering highlights
- ◆ PII scrubbing runs before any message reaches the LLM.
- ◆ Dedup checks on every inbound event keep webhook retries from creating duplicate records.
- ◆ Classification categories live in the database, so the business can change them without touching the workflow.
- ◆ A single reusable outbound-sender sub-workflow serves every alert and report.
Impact
- ✓ Management sees every open conversation and every SLA breach in one place.
- ✓ Unanswered customers are surfaced automatically instead of being discovered days later.
* Client names withheld under confidentiality. Data shown is anonymized or mock.
Next project WhatsApp AI Sales Concierge Music school · Indonesia