WhatsApp AI Booking Agent
An AI-assisted WhatsApp conversational workflow that answers common customer questions, collects project requirements, and books appointments into Google Calendar.
What is this project?
An automated WhatsApp booking assistant built with n8n and OpenAI. When a potential client sends a message on WhatsApp, the webhook triggers an intent recognition engine that answers frequently asked questions, extracts service preferences, and books discovery calls into Google Calendar.
What needed to be solved?
Small business owners lose valuable leads when inquiries arrive outside business hours or take too long to receive a response.
What was the intended outcome?
Build an automated conversational workflow that responds instantly on WhatsApp, collects lead requirements, and books qualified meetings without manual back-and-forth scheduling.
Execution & Implementation
MY ROLE: AI / Automation Builder
Configured WhatsApp Business Cloud API webhook integration via n8n.
Engineered an intent recognition prompt in OpenAI with strict guardrails to prevent hallucinated pricing or commitments.
Built two-way Google Calendar availability checking and event booking nodes.
Integrated error handling with fallback messages directing unresolvable queries to human review.
System & Process Workflow
Message Ingestion
Customer sends an inquiry via WhatsApp; webhook delivers the payload to n8n.
Intent Classification
OpenAI processes the message against service FAQs, business hours, and booking intent.
Slot Verification
If booking is requested, workflow checks Google Calendar for open 30-minute consultation slots.
Confirmation & Sync
Calendar invite is generated, and a confirmation message with meeting details is sent back.
Why I Built It This Way
Specific choices made during the project — and the reasoning behind each.
Strict AI Guardrails on Pricing
Engineered the OpenAI prompt to never discuss or estimate project pricing — always defer to human.
AI confidently quoting wrong prices to a potential client destroys trust before any conversation can happen. The bot qualifies, humans close.
n8n Self-Hosted (Not SaaS)
Used self-hosted n8n on a VPS rather than cloud-based automation platforms like Zapier or Make.
Self-hosted eliminates per-task pricing that grows with usage. At scale, task-based billing on SaaS platforms becomes expensive fast.
Human Fallback Before Escalation
Added a structured fallback message before any human escalation, giving the bot one more attempt to resolve via FAQ.
Escalating too quickly wastes human time. One graceful fallback attempt filters out simple questions that the bot missed on first pass.
Technologies & Software
Results & Impact
Interested in a Similar Project?
Let's discuss your goals, requirements, and timeline. I'll help you define the next step.
Start Project Discussion
Video Editing & Motion
Edited video content demonstrating structured pacing, narrative cutting, kinetic typography, multi-layered sound design, and color grading.