Skip to content
PM Master Handbook ⬇ PDF
Module 20

Automation with Zapier, Make and n8n

Connect Gmail, Slack, Google Sheets, forms and AI so routine work runs by itself, and learn to choose between Zapier, Make and n8n.

≈ 10 h studyDraftedSources: Zapier, Make and n8n docs and pricing pages; Slack API docs; Google OAuth docs; checked 23 Oct 2026

On this page
  1. The building blocks
  2. Choosing the tool
  3. How to connect Gmail and Slack (step by step)
  4. Zapier · Basic to advanced
  5. Make · Basic to advanced
  6. n8n · Basic to advanced
  7. Labs
  8. Reliability and safety
  9. Exercises with model answers
  10. Key takeaways
  11. Check your understanding
  12. Recommended resources

Use practice accounts

Automations act without asking. Build every lab with a practice Gmail and a practice Slack workspace, and never point a test at real customers. Screens are illustrative wireframes.

20.1 The building blocks

Every automation tool uses the same ideas with different names.

IdeaWhat it meansZapierMaken8n
AutomationThe whole flowZapScenarioWorkflow
TriggerWhat starts itTriggerTrigger module (first)Trigger node
ActionA step that does somethingActionModuleNode
ConditionOnly continue if…FilterFilter on a routeIF / Filter node
BranchDifferent paths for different casesPaths (paid)RouterSwitch / IF node
LoopRepeat for each itemLooping (paid)Iterator + AggregatorLoop Over Items
WebhookStart from another app's HTTP callWebhooks by Zapier (paid)Custom webhookWebhook node
ScheduleRun at set timesSchedule by ZapierSchedulingSchedule Trigger
AI stepAsk a model to classify, summarise or decideAI by Zapier / AgentsAI modulesAI Agent and chat model nodes
Billing unitWhat you pay forTasks (successful actions)CreditsExecutions (cloud); free self-hosted
Same idea, three names. Learn the ideas once, and you can use any of the three tools.

20.2 Choosing the tool

ZapierMaken8n
Ease for beginnersEasiestMedium (visual canvas)Medium to hard
Free tier100 tasks a month, two-step Zaps1,000 credits a month, 2 active scenarios, 15-minute minimum intervalCommunity Edition free to self-host (you pay the server)
First paid tierProfessionalCoreCloud Starter, about €20 to €24 a month
App integrations8,000+2,000+Hundreds of nodes plus any API via HTTP Request
Complex logicPaths, loops on paid plansStrong: routers, iterators, data storesStrongest: code, branches, sub-workflows
AI agentsZapier Agents, AI stepsAI modules and agentsAI Agent node with memory and tools; MCP
Data stays with youNo (cloud)No (cloud)Yes, if self-hosted
Best forQuick wins for non-technical teamsVisual, multi-branch flows on a budgetPower users, AI agents, high volume, privacy
Zapier vs Make vs n8n (checked 23 Oct 2026). A good path is Zapier for your first two automations, then n8n for anything with AI agents or high volume. Check each pricing page before paying.

20.3 How to connect Gmail and Slack (step by step)

20.3.1 Gmail

20.3.2 Slack

20.4 Zapier · Basic to advanced

Zapier Level 1

Build the Gmail → Slack Zap above. Then a second: Google Forms → Google Sheets (new response → add row).

Zapier Level 2 (paid)

  • Add a Filter (only continue if Subject contains "invoice").
  • Use Formatter to clean text, dates and numbers.
  • Add Paths for different cases (VIP vs normal).
  • Add an AI step (AI by Zapier) to summarise or classify before posting.

Zapier Level 3

  • Webhooks by Zapier to receive data from your Lovable or Replit app.
  • Zapier Agents for multi-step tasks described in plain language.
  • Zapier MCP gives Claude access to Zapier's app actions as tools.
  • Tables and Interfaces for small internal tools without a separate database.

20.5 Make · Basic to advanced

Make essentials

Router for branches, filters on each route, iterator + aggregator for lists (for example every attachment in an email), error handlers (resume, ignore, rollback) on risky modules, and data stores for memory between runs. Watch your credit use per run on the scenario's history page.

20.6 n8n · Basic to advanced

20.6.1 Getting n8n

Option How Cost
On your laptop Install Node.js, then run npx n8n and open http://localhost:5678 Free
Docker docker run -it --rm -p 5678:5678 n8nio/n8n Free
Small cloud server Docker on a small VPS or a free-tier VM (runs 24/7) Server cost only
n8n Cloud Sign up at n8n.io Paid plans after a trial

Claude.ai cannot reach localhost. For the MCP lab use n8n Cloud, a server, or a tunnel (for example Cloudflare Tunnel or ngrok's free tier).

20.6.2 AI agents in n8n

20.6.3 n8n and MCP

  • MCP Server Trigger node: turns a workflow into an MCP server that exposes the tool nodes attached to it (for example a Custom n8n Workflow Tool that runs another workflow). Add its production URL to Claude (Settings → Connectors → Add custom connector) or to Claude Code (claude mcp add --transport http …). Claude can then call your workflow as a tool.
  • MCP Client Tool node: lets an n8n AI agent use tools from other MCP servers.
  • Protect MCP and webhook URLs with authentication (header token), so only you can trigger them.

20.7 Labs

Lab 1 · Gmail → Slack in Zapier (Free, 15 minutes)

Build the two-step Zap shown above with a practice Gmail and Slack. Send yourself a matching email and check Slack within a few minutes.

Lab 2 · Form → Sheet → route in Make (Free, 30 minutes)

  1. Create a Google Form with Name, Email and Budget.
  2. Make: Watch responses → Router → Route A filter Budget > 50000 → Slack; Route B → Sheets add row.
  3. Submit two test responses and check both routes.

Lab 3 · AI email triage (n8n, 45 minutes)

  1. Gmail Trigger (practice account) → AI Agent with the Anthropic chat model.
  2. Instructions: "Classify the email as urgent, invoice, newsletter or other. Reply only with JSON: {category, sender, amount, due_date, summary}." Turn on structured output or add a Structured Output Parser.
  3. Switch on {{ $json.output.category }} → Slack for urgent; Sheets + Gmail label for invoices.
  4. Send four test emails, one per category. Check each lands correctly in Executions.

Lab 4 · Morning digest (n8n, 30 minutes)

Schedule Trigger 08:30 → Google Calendar (today's events) + Gmail (unread from the last 12 hours) → AI Agent "Write a 120-word brief with what to prepare" → Slack DM to yourself.

Lab 5 · Give Claude a tool (n8n Cloud or server, 30 minutes)

  1. Build a small sub-workflow: Execute Workflow Trigger (inputs title, severity) → Google Sheets append row → Slack message.
  2. In a second workflow add an MCP Server Trigger and attach a Custom n8n Workflow Tool that calls the sub-workflow. Name it log_bug and describe it and its two inputs clearly, because Claude reads that description.
  3. Turn on authentication in the trigger, activate the workflow and copy the production URL.
  4. Claude → Settings → Connectors → Add custom connector → paste the URL (add the token if asked).
  5. In a chat: "Log a bug: checkout button overlaps on small screens, severity high." Approve the tool call and check the sheet and Slack.

Lab 6 · Leads from your Lovable app (45 minutes)

Webhook node (copy its URL) → in your Lovable app (Class 16), ask Lovable to POST each new signup to that URL from a server function → n8n adds the lead to Sheets, asks Claude to score it, and alerts Slack for high scores. Class 21 adds a WhatsApp welcome message.

20.8 Reliability and safety

  • Test with sample data, then turn on. Keep the first week on "notify me" rather than "act".
  • Error alerts: in n8n set an error workflow that posts failures to Slack; in Make add error handlers; in Zapier turn on error notifications.
  • No duplicates: store an ID (email ID, form response ID) and skip items you've already processed.
  • Rate limits and costs: add waits for big batches; set API spend limits; watch task, credit and execution counts.
  • Privacy: only send the fields the AI step needs. Don't put customer personal data into tools your company hasn't approved.
  • Document each automation: what it does, owner, accounts used, how to switch it off.

20.9 Exercises with model answers

Exercise 1 · Design the flow

Every support email mentioning "refund" should create a row in a Refunds sheet and alert #support-leads if the amount is over ₹5,000. Sketch the nodes.

Model answer

Gmail Trigger (search "refund") → AI step extracts order ID and amount as JSON → Sheets append row → IF amount > 5000 → Slack alert to #support-leads. Add duplicate protection on the email ID and an error workflow.

Exercise 2 · Pick the tool

A non-technical ops team wants three simple automations this week. A growth team wants an AI agent processing 50,000 events a month and keeping data on company servers. Which tools?

Model answer

Ops team: Zapier (fastest to learn; free or Professional depending on steps). Growth team: self-hosted n8n (no per-task fees at high volume, AI Agent node, data stays on company infrastructure), with an engineer to run the server.

20.10 Key takeaways

  • Every automation is trigger → (check) → (AI) → action → notify.
  • Zapier for quick wins, Make for visual branching on a budget, n8n for AI agents, volume and privacy.
  • Connect Gmail and Slack with sign-in connectors in Zapier and Make; n8n self-hosted needs a Google OAuth app and a Slack bot token.
  • MCP Server Trigger turns an n8n workflow into a tool Claude can call.

20.11 Check your understanding

Check your understanding

What do you need to connect Gmail to self-hosted n8n?

  1. Only your Gmail password
  2. A Google OAuth client ID and secret from a Google Cloud project
  3. A Slack bot token
  4. Nothing
Why: Self-hosted n8n needs your own OAuth app; n8n Cloud, Zapier and Make provide one.

Which Slack scope does a bot need just to post messages?

  1. admin
  2. chat:write
  3. users:read.email
  4. files:write
Why: chat:write is enough to post; give bots the least access they need.

How can Claude call one of your n8n workflows?

  1. Email it the workflow file
  2. Use an MCP Server Trigger and add its URL as a custom connector
  3. Share your n8n password
  4. It can't
Why: The MCP Server Trigger exposes the workflow as a tool over MCP.

Why ask the AI step to answer in JSON?

  1. It's shorter
  2. Later nodes can reliably read fixed fields such as category
  3. Slack requires JSON
  4. It's cheaper
Why: Structured output makes branching and mapping predictable.