AI Automation
Built & Tested
2026
A Make.com workflow that reviews incoming Cal.com bookings, scores and routes each lead, logs the data, and prepares a personalized reply for human approval.
TOOLS USED
Make.com · Cal.com · Claude · Gmail · Google Sheets
HIGH-LEVEL WORKFLOW
01
Cal.com Booking
02
AI Evaluation
03
Score + Tier
04
Route
05
Notify + Log + Draft
HUMAN REVIEW
Nothing is sent to the prospect automatically. The reply is prepared as a Gmail draft.
PROBLEM
The repetitive first pass
A booking may contain everything needed to judge an opportunity, but someone still has to read it, understand the request, check fit, consider budget and region, record the information, and prepare a response.
The workflow was built to handle that repetitive first-pass work while keeping final communication under human control.
SYSTEM OVERVIEW
One evaluation, structured output
A Cal.com booking triggers a Make.com scenario as soon as it is submitted. The intake includes the lead’s name, email, website need, project details, country, and budget.
One Claude call evaluates the booking and returns structured fields: a 0–100 score, tier, intent, urgency, qualification status, a one-line summary, and a personalized draft reply. Parse JSON turns that response into fields Make can route and log.
AI EVALUATION OUTPUT
One evaluation returns the score, tier, intent, urgency, qualification result, summary and personalized reply.
HOW IT WORKS
From booking to reviewed draft
01 — A user submits the Cal.com booking form.
02 — Cal.com instantly triggers the Make.com scenario.
03 — One Claude call reviews the complete booking.
04 — Claude returns a score, tier, intent, urgency, qualification result, summary, and draft reply.
05 — Parse JSON converts the response into individually mappable fields.
06 — A router sends the lead through Hot, Warm, or Cold / Fallback paths.
07 — Each route sends a lead notification, creates a Gmail draft, and logs the lead in Google Sheets.
08 — Replies are never auto-sent; the Gmail draft waits for human review.
09 — Claude and JSON parsing both have dedicated error branches.
10 — If either step fails, an alert includes the raw booking so the lead remains available for manual handling.
WORKFLOW IN MAKE.COM
The full Make.com workflow: booking trigger, AI evaluation, structured parsing, tier routing, notifications, reply drafts and lead logging.
DECISION LOGIC
Business rules shape the route
Qualification uses project-specific business rules around budget, region, intent, urgency, and the detail supplied in the booking.
For this implementation, an out-of-region lead under $1k is treated as cold. An out-of-region lead at $1k+ can be treated as warm with a budget-verification flag.
These are working rules for this system, not universal qualification logic for every client. The underlying private prompt is intentionally not exposed.
Control stays visible. Failure stays recoverable.
HUMAN-IN-THE-LOOP
Draft first. Review before send.
Claude prepares the response, but Make creates it as a Gmail draft rather than automatically sending it.
The final communication stays under human review.
RELIABILITY / ERROR HANDLING
The raw lead survives a failed AI step.
Both realistic AI and formatting failure points have dedicated error paths. If Claude fails or Parse JSON fails, an alert email is generated with the raw booking details.
The practical guarantee is that a lead does not silently disappear because an AI step failed.
DRAFT FIRST
The personalized response is prepared as a Gmail draft. Nothing is automatically sent to the prospect.
IMPLEMENTATION NOTES
Inside the workflow
The system uses Cal.com for intake, Make.com for orchestration, one Claude call for evaluation and reply preparation, Parse JSON for structured fields, a router for tiered handling, Gmail drafts for controlled communication, and Google Sheets for the lead log.
The system was built and tested across the full intake path, from the booking trigger through evaluation, routing, draft preparation and lead logging.
GOOGLE SHEETS LEAD LOG
Each evaluated lead is recorded with the information needed for review and future handling.
WHAT IT DEMONSTRATES
Useful automation with visible judgment.
This project demonstrates a practical AI-assisted first pass: structured intake review, explainable business-rule routing, human approval before communication, and failure handling that preserves the original lead data.
It is an automation workflow with AI assistance—not an autonomous agent and not a claim of production results.
CURRENT STATUS
Built and tested. Not presented as live production.
Built and tested for my own business. Not yet running as the live production intake system.
Let’s find out what is actually worth automating.
Tell me where the work is getting repetitive, slow or difficult to manage.