Portfolio
Portfolio Assistant
Ask questions about the work, projects, or how to get in touch.
A production-grade multi-agent pipeline that watches a Google Form, routes each submission through four specialized AI agents — intake, pricing, scheduling, and communications — then fires a personalized Gmail confirmation automatically.
Project Overview
Business Problem
Small handyman businesses lose bookings because they can't respond fast enough. A customer fills out a form and waits — often for hours. This orchestrator eliminates that lag entirely: a quote, available time slots, and a confirmation email go out in under 60 seconds, automatically.
One-Sentence Summary
A no-code multi-agent n8n workflow that ingests handyman service requests from Google Forms and automatically returns a personalized quote, availability, and Gmail confirmation — zero human involvement required.
My Role
Designed the entire agent architecture from scratch — defined what each agent would do, wired together the orchestration flow in n8n, built the data-passing strategy between nodes, and connected the Google Sheets trigger and Gmail send.
Biggest Challenge
Passing context reliably between four separate agent webhooks without losing data. Each "Store" node acts as a state accumulator, re-attaching upstream results so downstream agents always have full context to work with.
What I Learned
Orchestration is harder than individual agents. Designing clean handoffs between specialized agents — and deciding which agent owns which decision — taught me to think about AI systems architecturally, not just prompt-by-prompt.
Tools Used
GitHub & Demo
Built entirely in n8n Cloud — no code files, so no traditional GitHub repo. The full workflow is exportable as a JSON file available on request. The workflow was demoed live as a Google Form submission that triggers the full 11-node pipeline and delivers a formatted Gmail estimate in under 10 seconds.
Interactive Workflow
Click any node to see what it does
Each node in the pipeline has a specific job. Click any step to learn what it does and why it matters.
How It Works
Step-by-step pipeline
1. Customer fills out Google Form
The form captures name, email, service address, description of services needed, and preferred dates/times. Responses land in a connected Google Sheet.
2. Google Sheets Trigger fires
n8n polls the sheet for new rows. The moment a submission appears, the workflow activates and extracts the four key fields into clean variables.
3. Intake Agent parses the request
The first AI agent reads the raw service description and structures it — identifying the service type, urgency, and key details that downstream agents will need.
4. Pricing Agent generates a quote
Using the structured job data, the pricing agent references the business's rate sheet and returns an itemized estimate with service type, complexity level, and final price.
5. Scheduling Agent finds availability
Checks the calendar against the customer's requested dates and returns 2–3 available appointment options formatted for the confirmation email.
6. Comms Agent drafts the response
Takes the pricing + scheduling results and composes a professional, personalized email — including the customer's name, service summary, quote, and time options.
7. Gmail sends the confirmation
n8n's Gmail node fires the email directly to the customer's address from the form — fully automated, no human touch required. Average end-to-end time: under 60 seconds.
Sample Output
What the customer receives
From: handyman@company.com
Subject: Service Request Update for Marcus
Key Features
🔗
Multi-Agent Architecture
Four specialized agents each own a distinct domain — intake, pricing, scheduling, and communication — keeping concerns separated and each agent's prompts tight and accurate.
🔄
State Accumulation Pattern
Each "Store" node re-attaches all upstream context before passing to the next agent. Downstream agents always have complete job + pricing + scheduling data without redundant API calls.
⚡
Fully Automated Trigger
The Google Sheets trigger polls continuously — no manual "run" step. The entire pipeline from form submission to customer email runs without any human intervention.
📊
Real Business Data
Connected to real Google Sheets for pricing lookups and a live Google Calendar for availability — not mocked data. This mirrors how a production deployment would actually operate.