// Music school · Indonesia
WhatsApp AI Sales Concierge
An AI customer-service agent that registers new leads and books trial lessons on WhatsApp, following the school’s own SOP.
- Client
- Music school · Indonesia
- Built at
- WithMi Automation
- Role
- Solution design & development
- n8n
- 5 workflows · 149 nodes
The problem
Every prospective student arrived through WhatsApp. Admins had to collect the same details by hand, log them in a guest book and juggle teacher availability to book trial lessons — which capped how many leads the team could handle.
The solution
A funnelling chatbot built as an AI agent with long-term memory and an FAQ knowledge base. It collects lead data conversationally, writes it to the guest book, notifies admins, then checks real teacher availability and books trial slots. A Redis debounce layer groups bursts of short messages into a single turn before the agent replies.
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.
67 nodes · Click a node to inspect it
Engineering highlights
- ◆ Redis debounce: customers who type in several short bursts get one coherent reply, not five.
- ◆ Structured output parser turns free-form chat into validated lead records.
- ◆ LLM-agnostic via OpenRouter, so the model can be swapped on cost or quality.
- ◆ Scheduled job resets teacher schedules so availability never drifts.
Impact
- ✓ The school can handle far more incoming leads than the admin team could manually.
- ✓ Leads are captured and followed up consistently, around the clock.
* Client names withheld under confidentiality. Data shown is anonymized or mock.
Next project Automated Assessment Report Engine Professional services group · US