A multi-channel virtual front desk for a small dental clinic, built in n8n as a single LangChain toolsAgent (the Dental Support Agent) wired to nine typed tools. The agent answers FAQs, books and cancels 30-minute appointments inside the clinic's Africa/Lagos business window, captures CRM data, escalates billing disputes and dental emergencies to humans on Slack, and replies in voice when a patient sends a Telegram voice note all through one shared Postgres-backed conversation memory.
The most impressive technical move is the single agent that owns the entire operation and reaches into the rest of the system through another workflow dispatcher that re-enters the workflow eight ways via a switch node.
0
28
I built an inbound sales automation flow that enriches and scores leads before CRM progression. The workflow pulls company context, updates HubSpot records, and applies intent checks so low-quality leads do not inflate pipeline stages.
1
34
I built an end-to-end sourcing flow from unstructured listing inputs to structured investment records. The system validates data against strict schemas, handles pagination and rate limits, and runs document checks on legal title PDFs to flag risks for human review.
0
39
I built one workflow that triages inbound Gmail into noise, operational, and meeting paths. Meeting requests are split into clear versus ambiguous before any calendar logic runs. Ambiguous requests get a clarify reply. Clear requests use real calendar availability plus booking caps from Postgres.
Key outcomes
- Prevents duplicate processing with DB-backed dedupe
- Uses confidence thresholds to avoid low-quality sends
- Keeps outbound mail plain text and auditable
- Supports exact-match auto-booking only when conditions are safe