← Content Library (LinkedIn Growth Mastery)All programsHomeSearch
Content Library (LinkedIn Growth Mastery)·Bonus Sessions - AI Automations Mastery·03:14:25

AI Automation Mastery I: Trigger-Logic-Action, Zapier vs Make vs n8n, and Four Live n8n Builds (Form to Email, AI-Personalised Email, Sheets and Data Tables, Scheduled LinkedIn Poster)

Jitesh Duggal Trainer - AI automation agency founder; Growth School / Outskill AI educator · Kartik Host (Outskill) - logistics, Slido Q&A, the AI Generalist roadmap and fellowship segment
▶LMS lesson
← PreviousLinkedIn Sales Navigator: Free vs Premium vs Navigator, Personas, Filtering 5,000,000 CEOs
Lesson 1 of 2 in Bonus Sessions - AI Automations Mastery
Next →AI Automation Mastery II: an AI LinkedIn Pipeline with Human-in-the-Loop Approval and Nano

The short version

  1. Every automation is a trigger, optional logic, and an action - taught through a pizza-delivery analogy; triggers are scheduled, manual or app-based (0:24-0:48).
  2. Zapier (6,000+ apps, cloud only, expensive), Make (visual, cheaper), n8n (open source, cloud or self-host ~$5-6/month VPS, 1,500+ apps, AI-native). Start on n8n's 14-day cloud trial; a node is a step, a credential is the permission (0:48-1:08).
  3. Builds: a feedback form that emails a thank-you (Gmail, drag-to-map fields, attribution off, publish for the production URL); an OpenAI node that writes the email - API key not ChatGPT subscription, GPT-4.1 nano, iterate the prompt, pin data; log to Google Sheets, then n8n Data Tables when simultaneous writes overwrote rows; a 9 AM scheduled LinkedIn poster reading a 'Ready' row and marking it 'Posted' (1:08-2:33).
  4. Q&A: webhooks for websites, bulk email and Gmail limits (~500/day personal, ~2,000 Workspace), Outlook/SMTP, LinkedIn automation risk, Telegram vs WhatsApp, sub-workflows, cloud vs self-host, JSON export to GitHub, 9,600+ templates (2:43-3:14).

At a glance, three clicks deep

Skim here first: the closed row is the glance, open is the study card with the key points and timestamps, and the ↓ link drops to that concept's full write-up below.

01Trigger, logic, action - and the three kinds of triggerTrigger + optional logic + action;›

Trigger + optional logic + action; scheduled / manual / app triggers.

Pizza analogy (0:24-0:33)

Framework (0:35-0:39)

Trigger types and three cases (0:39-0:48)

↓ Full write-up of this concept

02Zapier vs Make vs n8n, and cloud vs self-hostingn8n on cloud trial first;›

n8n on cloud trial first; VPS later.

Comparison (0:48-0:53)

Open source and VPS (0:53-1:01)

Nodes and credentials (1:03-1:08)

↓ Full write-up of this concept

03Build 1: feedback form -> Gmail thank-youForm trigger -> Gmail send;›

Form trigger -> Gmail send; map by dragging; publish for the live URL.

Form node (1:08-1:16)

Gmail mapping and options (1:16-1:31)

Executions (1:33-1:36)

↓ Full write-up of this concept

04Build 2: an OpenAI node writes the email; model tiers and API keysAPI key credential;›

API key credential; nano-tier model; iterate prompt; pin data.

Key and credits (1:38-1:41)

Prompt iteration and pin data (1:41-1:56)

Model tiers (1:46-1:50)

↓ Full write-up of this concept

05Build 3: log to Google Sheets - then n8n Data Tables when rows overwriteSheets for convenience;›

Sheets for convenience; Data Tables for concurrent writes.

Append Row (2:02-2:05)

Overwrite bug (2:05-2:07)

Data Table fix (2:07-2:09)

↓ Full write-up of this concept

06Build 4: the 9 AM scheduled LinkedIn posterSchedule -> first Ready row -> LinkedIn post -> mark Posted.›

Schedule -> first Ready row -> LinkedIn post -> mark Posted.

Sheet seeding (2:10-2:14)

Get Rows + first match (2:14-2:21)

LinkedIn credential workaround (2:21-2:26)

Update Row and live run (2:26-2:33)

↓ Full write-up of this concept

07Q&A: webhooks, email limits, LinkedIn risk, sub-workflows, JSON exportWebhooks;›

Webhooks; Gmail limits; official node only; sub-workflows; JSON export.

Webhooks (2:44)

Email limits and providers (2:50-2:56)

LinkedIn risk, Telegram/WhatsApp (2:54-2:58)

Sub-workflows and export (2:59-3:08)

↓ Full write-up of this concept

The concepts in full

01

Trigger, logic, action - and the three kinds of trigger

Order the pizza (trigger). Is the shop open, who's closest (logic). Cook it, text the customer (action).

Every automation has at least one trigger and one action; logic is optional. Triggers: scheduled (the alarm clock), manual (you click send), app-based (a new Gmail, a Zoom join, a LinkedIn message). Three worked cases: new Sheet row -> is it scheduled today? -> post to LinkedIn; website form -> validate email -> send the PDF; newsletter signup -> no logic -> five-email series. Plan the three parts before touching the canvas.

Why it matters

The planning model every later build uses.

This idea elsewhereparallelBC9's n8n anatomyparallelThe mastermind's Make.com trigger/action
02

Zapier vs Make vs n8n, and cloud vs self-hosting

Open source means the code is on GitHub and you may run it yourself for free - but start on the trial, not the server.

Zapier: 10+ years, beginner-friendly, 6,000+ integrations, cloud only, expensive. Make: drag-and-drop canvas, complex flows, cheaper, cloud only. n8n: open source (repo shown live), cloud or self-hosted, code when needed, flat pricing, 1,500+ integrations, AI-native. Self-host options: your own machine (only runs while it is on) or a VPS from ~$5-6/month; the recommendation is the 14-day cloud trial with no card. Cloud adds 'Build with AI' and one-click Google sign-in credentials that self-hosted lacks. Nodes are steps; credentials are OAuth or API-key permissions; n8n cited as GDPR and SOC 2 compliant.

Why it matters

Choose the platform once and stop relitigating it.

This idea elsewhereparallelBasecamp 3's case for n8nparallelBasecamp 3 on deployment economics
03

Build 1: feedback form -> Gmail thank-you

Drag the email field into 'To', switch off the n8n footer, publish, and hand the class the URL.

PROCEDURE: On Form Submission trigger - title 'Feedback form', fields Name (text), Email (email), Feedback Rating (dropdown Good/Okay/Bad), Feedback Notes; Execute Step opens the form. Gmail 'Send a message' - Sign in with Google; drag the form's email output into To (expression mode); subject and body typed with the name dragged in; Options -> 'Append n8n attribution' off; Sender Name set. Publish -> production URL (test URL only works before publishing). Review runs in Executions; deactivate to stop.

Why it matters

The smallest complete automation, and every UI habit you need.

04

Build 2: an OpenAI node writes the email; model tiers and API keys

Your ChatGPT subscription cannot run inside an automation. A $5 API credit can.

Insert OpenAI 'Message a Model' between form and Gmail. API key from platform.openai.com -> API Keys -> Create secret key -> n8n credential; billing is prepaid credits from ~$5, separate from ChatGPT Plus. Model: GPT-4.1 nano (GPT-5 nano was slower for him at similar cost). Prompt in expression mode with the form fields dragged in; refine on bad output - no preamble, no duplicate subject, concise, no '[Name]' placeholders, fixed 'Best regards' signature. Pin Data on the trigger to reuse one submission while iterating. Re-map Gmail body to the model's output; republish. Tiers: OpenAI flagship -> mini -> nano; Anthropic Opus -> Sonnet -> Haiku; Gemini Pro vs standard - cheapest tier for simple writing.

Why it matters

The first AI step, and the cost discipline that keeps it cheap.

This idea elsewhereparallelBC9's API-vs-subscription cost mathparallelBasecamp 3's model-sizing rule
05

Build 3: log to Google Sheets - then n8n Data Tables when rows overwrite

Thirty people submitted at once and Google Sheets wrote them all into the same row.

Google Sheets 'Append Row' mapped from the form (Google OAuth, no separate API). Live failure: near-simultaneous submissions overwrote each other - a known Sheets API race under concurrent writes. Fix: n8n's native Data Table - Create Data Table with columns Name/Email/Rating/Notes, then 'Insert Row' pointed at it; no external API, no overwrite.

Why it matters

A real concurrency bug and the built-in cure.

06

Build 4: the 9 AM scheduled LinkedIn poster

Claude drafts ten posts into a sheet; every morning n8n takes the first 'Ready' row, posts it, and marks it 'Posted'.

PROCEDURE: a Google Sheet with Post Text and Status (Ready/Posted), seeded with Claude-drafted posts. Schedule Trigger daily 9:00. Google Sheets 'Get Rows' filtered Status = Ready, Options -> 'Return only first matching row'. LinkedIn 'Create a Post' - credential bug workaround: Set up manually, disable Organization Support, connect, switch Organization back to Person so the Person field populates; map Post Text; pin the Get Rows data. Google Sheets 'Update Row' matched on row number -> Status = Posted. Execute Workflow live: row fetched, posted, marked. The official LinkedIn node cannot comment or reply to DMs. Formatting of the first post was poor and fixed in the sheet.

Why it matters

The pattern for any queue-driven publisher.

This idea elsewherecontinuesDay II adds AI generation, images and human approvalparallelThe content calendar this could publish
07

Q&A: webhooks, email limits, LinkedIn risk, sub-workflows, JSON export

Any website can start a workflow if it calls the webhook URL. Any workflow can be a file on GitHub.

Webhook trigger replaces the form for WordPress/Wix/Shopify/custom sites - the site calls n8n's webhook URL on the event. Bulk personalised email is fine but Gmail caps ~500/day personal, ~2,000 Workspace (figures the trainer was unsure of); Outlook or generic IMAP/SMTP nodes as alternatives; Microsoft 365 Excel integrates, local Excel does not. LinkedIn: posting via the official node is allowed with minimal risk; third-party automation (PhantomBuster, Typefully and others) carries ban risk. Telegram is a 5-10 minute trigger; WhatsApp needs Meta business verification. 'When executed by another workflow' + 'Execute Sub-workflow' nests workflows. Export: workflow menu -> Download JSON (no credentials) -> import elsewhere and reconnect; store in GitHub. Templates: 9,600+, 127+ tagged LinkedIn. Typical API spend $5-30.

Why it matters

The operational questions every first-week builder asks, answered in one place.

This idea elsewhereparallelSession 17's webhook plumbing

Tools referenced

ToolCoverageMomentContext
n8ndemonstratedAll four builds; cloud trial
Gmaildemonstrated
OpenAI PlatformdemonstratedAPI key; GPT-4.1 nano
GPT-4.1demonstratednano tier
Google Sheetsdemonstrated
LinkedIndemonstratedOfficial n8n node
ClaudedemonstratedDrafted the posts
GitHubdemonstratedn8n repo; JSON storage
Slidodemonstrated
Zapierexplained
Make.comexplained
Google Geminimentioned
DeepSeekmentioned
Telegrammentioned
WhatsAppmentioned
Outlookmentioned
PhantomBustermentionedBan-risk caution
Typefullymentioned
SupergrowmentionedDelayed course tool

Session materials

Archived locally on V: — click to open. Companion pages link to the LMS.

lms
AI Automation Mastery - II
companion page on the platform

Action items

    Resources mentioned

    Resources
    • docFeedback-form workflow JSON
    • docRecording + transcript same day; whiteboard, post-read and Office Hour 1 workbook after Day II

    Extraction notes

    This page was built from an auto-generated transcript, which garbles product and people's names. Those were corrected silently in everything above and logged here for transparency. The warnings flag claims that were true on the recording day but change fast.

    Transcript corrections applied

    The transcript saysThe trainer actually means
    MediaJadetrainer's agency name - unverified
    Harisha third-party LinkedIn tool - unresolved
    Humanaeck / Humanaea course tool code - unresolved
    an Aiden / any 10 / n a 10 / an 8 10n8n
    Charge GPT / ChargegbtChatGPT
    wipe codingvibe coding

    True on recording day — verify before relying

    ← PreviousLinkedIn Sales Navigator: Free vs Premium vs Navigator, Personas, Filtering 5,000,000 CEOsNext →AI Automation Mastery II: an AI LinkedIn Pipeline with Human-in-the-Loop Approval and Nano