Diagnose manual business frictions, calculate time savings ROI, explore battle-tested open-source repositories, and deploy 5-phase production prompt chains.
Monitors inbox, classifies incoming intents, drafts human-in-the-loop responses in Drafts folder, and checks Google Calendar for available slots.
Pre-built Gmail, Calendar, and LLM pieces with self-hosted privacy guarantees.
Ready-to-import blueprint JSON with filter routers and human-in-the-loop draft nodes.
Filter unread business emails, extract sender, subject, body & thread metadata.
Classify into Meeting/Approval/Inquiry/Spam. Query Calendar API for free slots if meeting.
Archive spam, route standard inquiries to drafts, and escalate VIPs to Slack.
Create email draft without sending, allowing 1-click human review and dispatch.
You are a Principal Enterprise Automation Architect. I am an Executive Assistant looking to build an automated email and calendar workflow using Make.com (Integromat) + LLM. Business Requirements: 1. Automatically read incoming business emails and classify intent (Meeting request, inquiry, approval, or spam). 2. Generate draft replies strictly within Gmail Drafts folder for human review (NEVER send autonomously). 3. If a meeting is requested, check Google Calendar for available slots and propose 3 options. Please provide: The complete Make node diagram, module configurations, and why the Human-in-the-loop protocol is mandatory.
Write the System Prompt for the LLM node in Make.com/n8n. Enforce strict JSON output matching this schema:
```json
{
"category": "meeting | approval | general_inquiry | spam",
"urgency": "low | medium | high",
"summary_one_sentence": "One-sentence summary of sender's core request",
"requires_calendar": true,
"reply_draft_subject": "Re: Original Subject",
"reply_draft_body": "Polite business response with 3 proposed time slots",
"action_items": ["Action 1", "Action 2"]
}
```
Ensure valid, parseable JSON without extra conversational preamble.Using everyday office metaphors, explain these 4 core automation concepts to non-technical colleagues: 1. What is a Webhook / Trigger? (Metaphor: The front desk doorbell) 2. What is Structured Output (JSON)? Why shouldn't LLMs output freeform prose? 3. What is Human-in-the-loop? Why is it the lifeblood of enterprise safety? 4. What is an API Token / OAuth permission?
My automation failed at the "Create a Draft" node. Act as a senior debugging expert and diagnose the top 4 common failure points: 1. Message ID vs. Thread ID mapping mismatch breaking conversation threads. 2. Unescaped markdown backticks or newline characters causing JSON parse errors. 3. Missing 'gmail.compose' OAuth scopes on the Google Workspace account. 4. Filtering out automated 'noreply@' addresses before calling the LLM.
Create a Team Standard Operating Procedure (SOP) for this Email & Calendar automation: 1. Daily EA Routine: Reviewing Drafts in batch, verifying proposed slots, and 1-click dispatch. 2. Failover Protocol: Fallback steps if the LLM provider experiences latency or rate limits. 3. Privacy & Compliance: How corporate data is encrypted in transit and credential isolation.