Unit V · Applications and the Future

20. AI workflow automation

A workflow is a series of steps to finish a task. Automation means the steps happen by themselves, without a person clicking each one. AI workflow automation means one or more of those steps are done by AI, and the output of one step becomes the input of the next.

This covers Lab experiments 18 and 19.

A simple example

Task: Every week, make a short newsletter for the class.

Manual workflow:

  1. Collect the week's notices. (You)
  2. Write a summary. (You)
  3. Make a nice image. (You)
  4. Post it on the WhatsApp group. (You)

Automated workflow:

  1. Notices arrive in a Google Form. (Automatic)
  2. AI summarises them into a newsletter. (AI)
  3. AI makes an image. (AI)
  4. The newsletter is emailed to the class. (Automatic)

You set it up once. Then it runs every week.

Chain of tools (the lab way)

The simplest automation is a chain: the output of tool 1 goes into tool 2, and so on. In the lab you can do this by hand, to understand it.

Experiment 19: Document → Summary → Presentation

  1. Create a document. Ask Gemini to write a 2-page report on "Electric buses for Visakhapatnam". Paste it into Google Docs.
  2. Summarise. Give the document to NotebookLM or ChatGPT. Ask for a 10-point summary.
  3. Present. Paste the summary into Gamma using "Paste in text". Get a 10-slide deck.
  4. Extra steps. Ask ElevenLabs for narration. Ask Canva for animation. Ask AI for a social media post about the deck.

Each tool's output is the next tool's input. That is a workflow. Write down each step and the time it took.

Automation tools (no code)

These tools connect apps together. "When X happens, do Y, then Z."

ToolLinkNotes
Zapierzapier.comEasiest. Free plan with limits. Has AI steps built in.
Makemake.comVisual, drag-and-drop. Free plan.
n8nn8n.ioOpen source. Can run on your own laptop. Strong AI features.
Google Apps Scriptscript.google.comFree. Automates Gmail, Sheets, Docs with small code. Gemini can write the code.
IFTTTifttt.comSimple "if this, then that" rules.

Experiment 18 example with Zapier or Make:

  • Trigger: A new row is added to a Google Sheet (a student's question).
  • AI step: Send the question to ChatGPT or Gemini with a prompt: "Answer this in simple English in under 50 words."
  • Action: Write the answer in the next column of the sheet. Or send it by email.

Many such tools have a free plan with 100 runs a month. That is enough for the lab.

Automation with a little code

With an API key (Chapter 19) and Python, you can write your own automation. Example: a Python script that reads all .txt files in a folder, asks Gemini to summarise each one, and saves a summary.txt. Ask an AI to write this script for you.

Google Apps Script is also powerful. Ask Gemini: "Write a Google Apps Script that, for every row in my sheet, sends column A to Gemini and writes the answer in column B."

AI agents

The newest kind of automation is the AI agent. An agent is an AI that can plan steps and use tools on its own. You say the goal. It decides the steps. Examples: Gemini's and ChatGPT's agent modes, Claude Code, OpenAI Operator, Manus. Agents are new and still make mistakes, so a human must check the result.

Things to remember

  • Start by doing the workflow by hand once. Then automate it.
  • Keep a human check at the end for anything important.
  • Automation that sends emails or messages can make mistakes at scale. Test with yourself first.
  • Do not automate anything that needs personal data without permission.

Try it

  • Do the Document → Summary → Presentation chain above with 3 different tools. Write down each step.
  • Sign up at zapier.com or make.com. Make a 3-step zap: Google Form → AI summary → Gmail. Test it with yourself as the receiver.
  • Open script.google.com and ask Gemini to write a script that emails you a joke every morning.
  • Watch: n8n in 100 seconds — search for a short beginner video.

New words

  • Workflow — a series of steps to complete a task.
  • Automation — making steps happen without a person doing each one.
  • Chain — a line of tools where each one's output feeds the next.
  • Trigger — the event that starts an automation ("a new form response").
  • Action — a step the automation does ("send an email").
  • Zap / scenario — one automation in Zapier / Make.
  • Open source — software whose code is free to see, use and change.
  • Agent — an AI that plans its own steps and uses tools to reach a goal.
  • At scale — in large numbers; a mistake repeated many times.