# ChatGPT: Turn a Manual Process into an Automation Plan Before You Build

> Have ChatGPT map a manual process into rule and judgment steps, answer its questions, then get the smallest n8n, Make or Zapier plan and tests.

Source: https://learn.ynteractive.com/content/chatgpt-plan-automation-from-manual-process · Updated 2026-11-17 · Free from Stepthrough (https://learn.ynteractive.com)

## Practice it step by step

[Start the free practice](https://learn.ynteractive.com/modules/chatgpt-plan-automation-from-manual-process)

[Take the free module: why it matters, practice and a quick check →](https://learn.ynteractive.com/courses/plan-and-fix-automations-with-chatgpt/1)

You do every step yourself in a practice copy of ChatGPT. It checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, about 9 minutes, and nothing touches your real accounts. [Open full screen](https://learn.ynteractive.com/modules/chatgpt-plan-automation-from-manual-process)

Free · Live now · [ChatGPT for Everyday Work](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work) · [ChatGPT](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work) · ~9 min · 20 steps · Beginner

Updated November 17, 2026. Built against the tool’s current screens and checked step by step. [How we check this lesson](https://learn.ynteractive.com/content/about#how-we-check)

Before you open [n8n](https://learn.ynteractive.com/content/tool/n8n), Make or [Zapier](https://learn.ynteractive.com/courses/ai-automation-with-zapier), write down how the job is done today and let [ChatGPT](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work) take it apart. In this free, interactive lesson you attach an office manager's own description of her Friday timesheet chase, have ChatGPT map every step and sort rules a computer can follow from judgments a person should keep, answer its questions, and then ask for the smallest [automation](https://learn.ynteractive.com/content/glossary/automation) plan, with tests and what to leave manual. The agency is made up.

## At a glance

| Question | Answer |
| --- | --- |
| What you build | Attach a written-down manual process, get every step sorted into rules a computer can follow and judgments a person keeps, answer the open questions, then get the smallest automation plan, what to test and what to leave manual. |
| Tools | ChatGPT |
| Time | About 9 minutes, 20 steps in 4 parts |
| Level | Beginner, no coding needed |
| Cost | Free |
| Ways to learn it | [Practice it step by step](https://learn.ynteractive.com/modules/chatgpt-plan-automation-from-manual-process) |

### The 4 parts, in short

1. Start from how it is done today: Read how she does it now, Turn on Temporary chat, Open the + menu, Add photos & files, and 2 more steps.
2. Map every step first: Click the message box, Ask for the step map, Send it.
3. Sort rules from judgment: Read the step map, Notice what it would not call a rule, Read the six questions to answer first.
4. Ask for the smallest plan: Click the message box, Answer and ask for the plan, Send it, Read the six automations, and 4 more steps.

**Start here first:** [Summarize a Meeting Transcript with ChatGPT](https://learn.ynteractive.com/content/chatgpt-summarize-meeting-transcript) (module 1 of ChatGPT for Everyday Work)

**In the course:** [Part 3 · ChatGPT for Everyday Work](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work) (Module 40 of the [full course](https://learn.ynteractive.com/content/course))

## The business case: automations break on the step nobody wrote down

Teams often start building straight away: a schedule, a spreadsheet lookup, a batch of emails. The trouble comes later, from the cases nobody described: a worker who submits twice with different hours, a supervisor who approves on an old [email thread](https://learn.ynteractive.com/content/glossary#email-thread), a correction someone keeps in their head.

Each of those turns into a payroll mistake once a machine does the job without asking. Not sure whether a job needs an automation or an [AI agent](https://learn.ynteractive.com/content/n8n-ai-agent-tutorial-google-sheets)? See [How to Get ChatGPT in Agent Mode: AI Agent vs Chatbot vs Automation](https://learn.ynteractive.com/content/ai-agent-vs-chatbot-vs-automation). Planning first costs ten minutes. Rebuilding a live workflow, or paying people late, costs much more.

What changes: ChatGPT turns a plain description into a step table, marks which steps are rules and which are judgment, lists the questions to answer, and only then writes a plan with a trigger, an approval point and a "never do this by itself" line for each automation.

## The example: a staffing agency's Friday timesheet chase

Northgate Staffing (made up) places about 60 temporary workers at 14 client sites and runs payroll every Monday. On Friday about 20 timesheets are missing. The office manager emails each worker, texts some from her own phone, calls the last few at noon, emails 14 supervisors for approval, chases 4 of them again at 3 PM, and sends payroll a list at 4:30 PM. Three workers were paid a week late last month, and twice she paid the wrong one of two submissions.

Download the file and follow along: [friday-timesheet-chase.txt](/content/files/chatgpt-plan-automation-from-manual-process/friday-timesheet-chase.txt).

## Who this lesson is for

Office managers, operations leads and anyone asked to "automate this" who does not write code. If you have never built an automation, start with [your first AI automation](https://learn.ynteractive.com/content/first-ai-automation-what-to-automate-first-and-how-to-test) or the [Intro to AI Automation course](/courses/intro-to-ai-automation). This lesson is the first of three in [Plan and Fix Automations with ChatGPT](/courses/plan-and-fix-automations-with-chatgpt).

## What ChatGPT produced in the real run

The first answer split the process into 15 steps and marked the noon calls and hour corrections as Human, the step where she picks who to text as Mixed, and the rest as Rule-based. It listed 34 questions and named six to start with, such as how workers are uniquely identified and what happens when a worker submits twice. It said no workflow or platform had been proposed yet.

The second answer picked n8n with [Google Sheets](https://learn.ynteractive.com/courses/gemini-jump-start), Forms and Gmail, and said Make or [Zapier](https://learn.ynteractive.com/content/glossary#zapier-zap) could also handle much of it. It gave six automations plus a supporting workflow for approval-link responses, a 10-item test checklist with a "shadow mode" week, and a table of seven things to leave manual. Asked for the smallest version, it still recommended building seven workflows. The lesson's advice is to build only the Friday 9 AM reminder first.

Nothing in this lesson was built or run inside n8n, Make or Zapier. The plan is ChatGPT's written plan, not a tested workflow.

## What this saves you

- Rework: you find the duplicate and approval rules before they are baked into a live workflow.
- Payroll mistakes: the steps that need a person are named before anything sends by itself.
- Time: a clear first build (one reminder) instead of a week spent on seven workflows.
- What it does not replace: testing on a copy of the sheet with test people, and the office manager's say on what stays manual.

## Gotchas

- Say "don't design anything yet" in the first prompt, or you get a build plan before the rules are clear.
- Ask for "rule or judgment" on every step. A step nobody can state as a rule is not ready to automate.
- Answer its top questions in your next prompt. The plan is only as good as those answers.
- "Smallest version" can still come back big. Pick the one automation with no approvals and the lowest risk, and build that first.
- Never test with real workers or real emails. Use a copy of the sheet and a test Gmail account.
- The file names real-looking people and contact details: use a Temporary Chat, or your company's approved AI tool.

## Next business case

Adding an AI step to your plan? [Write and test the prompt for an AI step](https://learn.ynteractive.com/content/chatgpt-write-test-ai-step-prompt-n8n-make-zapier). Something you did not build stopped working? [Read an automation error and fix it](https://learn.ynteractive.com/content/chatgpt-fix-automation-error-message). Ready to build the reminder? Pick a tool with [n8n vs Make](https://learn.ynteractive.com/content/n8n-vs-make-which-should-my-team-use), then start with [your first n8n workflow](https://learn.ynteractive.com/content/n8n-tutorial-for-beginners-first-workflow), [your first Make scenario](https://learn.ynteractive.com/content/make-com-google-forms-to-google-sheets) or [your first Zap](https://learn.ynteractive.com/content/zapier-google-forms-to-google-sheets).

## Try it yourself, free

Reading the steps is a start. Doing them is how it sticks. The interactive module walks you through every click in a practice copy of ChatGPT, checks each step, and never touches your real accounts.

[Start “ChatGPT: Turn a Manual Process into an Automation Plan Before You Build”](https://learn.ynteractive.com/modules/chatgpt-plan-automation-from-manual-process)

## Step by step: ChatGPT: Turn a Manual Process into an Automation Plan Before You Build

All 20 steps of the interactive module, in order. In the module you watch each one, then do it yourself in a practice copy that checks your work.

### 1. Start from how it is done today

1. **Read how she does it now.** Every Friday the office manager spends hours chasing **20 missing timesheets** and **14 supervisors**. Three workers were paid late last month. Read her own description, then click **✕**.
   Every Friday, the office manager chases twenty missing timesheets and fourteen supervisors. Three workers were paid late last month. Read her own description, then close the file.
2. **Turn on Temporary chat.** This file has worker and supervisor details. Click the **Temporary chat** icon at the top right so it stays out of your history.
   This file has worker and supervisor details.
3. **Open the + menu.** Click the **+** next to the message box.
4. **Add photos & files.** Click **Add photos & files**.
5. **Pick friday-timesheet-chase.txt.** Pick **friday-timesheet-chase.txt**.
   Pick the timesheet file.
6. **Open it.** Click **Open**. ChatGPT will plan from **her** words, not a generic idea of payroll.
   ChatGPT will plan from her own words, not a general idea of payroll.

### 2. Map every step first

7. **Click the message box.** The habit: **don’t design anything yet**. First get every step in a table, sorted into **rules a computer can follow** and **judgments she should keep**, plus the questions to answer before building. Click the **message box**.
   Here's the habit. Don't design anything yet. First, get every step in a table. Sort each one into a rule a computer can follow, or a judgment she should keep. Then the questions to answer before building.
8. **Ask for the step map.** Paste in the request. It names the tools (a no-code tool and **Google Workspace**), asks for trigger, action, information and **rule or judgment** for each step, then the open questions.
   This names the tools, asks for a table of every step, rule or judgment for each one, and the questions to answer first.
   ```text
   Attached is how our office manager chases timesheets every Friday, in her own words. We want to automate the repetitive parts with a no-code tool (n8n, Make or Zapier) and Google Workspace. Don't design anything yet. First turn her description into a table of every step: what starts it, what she does, what information it needs, and whether it is a rule a computer can follow or a judgment she should keep. Then list the questions you would need answered before building.
   ```
9. **Send it.** Click **Send** (the arrow on the right).

### 3. Sort rules from judgment

10. **Read the step map.** Fifteen steps from her seven paragraphs. Most are **Rule-based**. The noon calls (7) and the hour corrections (11 and 12) are **Human**, exactly what she said she wants to keep. Click the table.
    Fifteen steps from her seven paragraphs. Most are rules. The noon phone calls, and the hour corrections, are marked human. That's exactly what she said she wants to keep.
11. **Notice what it would not call a rule.** Step 5 is **Mixed**: nobody wrote down how she picks the eight people she texts. A rule nobody can state is not a rule yet. Click the line.
    Step five is mixed. Nobody wrote down how she picks the eight people she texts. A rule nobody can state is not a rule yet.
12. **Read the six questions to answer first.** Out of **34** questions, these six come first: who is who, what to do with **two submissions**, what counts as an **approval**, and the **real payroll cutoff**. You answer them in the next prompt. Click the questions.
    Out of thirty four questions, these six come first. How to tell workers apart. What to do when someone submits twice. What counts as an approval. And the real payroll cutoff. You answer them next.

### 4. Ask for the smallest plan

13. **Click the message box.** Now answer its questions, and ask for the **smallest version worth building first**. Click the **message box**.
    Now answer its questions, and ask for the smallest version worth building first.
14. **Answer and ask for the plan.** Paste in the request. Email only, cutoff **Monday 10 AM**, supervisors **click a link**, duplicates keep the **latest entry but flag it**. Then one trigger per automation, who approves, what it must **never** do, what to test and what to leave manual.
    This gives four answers. Email only. Payroll closes Monday at ten. Supervisors click a link. Duplicates keep the latest entry, but get flagged. Then it asks for the plan, the tests, and what to leave manual.
    ```text
    Answers: reminders can go by email only for now; payroll cutoff is Monday 10 AM; supervisors can click a link instead of replying; and duplicates should keep the latest form entry but flag it. Now write the automation plan: for each automation, one trigger, the steps in order, where a person approves, and what it must never do by itself. Keep it to the smallest version worth building first, tell me what to test before it goes live, and what to leave manual.
    ```
15. **Send it.** Click **Send** (the arrow on the right).
16. **Read the six automations.** One trigger each, who approves, and what it must **never do by itself**, like treating silence as approval. It chose **n8n** but says Make or Zapier could do it too. Click the table.
    One trigger each, who approves, and what each one must never do by itself. Like treating silence as approval. It picked n eight n, but says Make or Zapier could do it too.
17. **Read the Friday reminder steps.** Automation 2 in six steps: find who has not submitted, send **one** reminder, and list anyone whose email failed. No approval needed. Keep this one in mind. Click the steps.
    Automation two, in six steps. Find who hasn't submitted. Send one reminder. List anyone whose email failed. No approval needed. Keep this one in mind.
18. **Read what to test before it goes live.** Test on a **copy of the sheet** with test people, never real workers. The duplicate test (40 hours, then 32) checks the exact mistake that cost her money. Then one full Friday in **shadow mode**, sending nothing. Click the tests.
    Test on a copy of the sheet, with test people, never real workers. The duplicate test checks the exact mistake that cost her money. Then one full Friday in shadow mode, sending nothing.
19. **Read what stays manual.** The calls, the hour corrections, the conflicts and the final payroll decision stay with people. Automating these is where a payroll mistake would hide. Click the table.
    The phone calls, the hour corrections, the conflicts, and the final payroll decision stay with people. That's where a payroll mistake would hide.
20. **Check the scope yourself.** You asked for the **smallest version**, and it still says build **seven** workflows. That is a lot for an office where nobody writes code. Our advice: build **only the Friday 9 AM reminder** first. It needs no approvals and has the lowest risk. Once it works, add the next one. Click the recommendation.
    You asked for the smallest version, and it still says build seven workflows. That's a lot for an office where nobody writes code. Our advice. Build only the Friday morning reminder first. It needs no approvals, and has the lowest risk. Once it works, add the next one. You mapped a manual process, and planned the automation, before building anything.

## Frequently asked questions

**Can ChatGPT help me plan an automation?**

Yes. Write down how the job is done today, attach it, and ask ChatGPT to map every step and mark each one as a rule or a judgment before it designs anything. Then answer its questions and ask for the smallest plan, what to test and what to leave manual.

**How do I know which parts of a process to automate?**

Automate the steps a computer can follow from a rule you can write down, like "email everyone with a blank Hours column at 9 AM". Keep steps that need judgment, like accepting a correction to someone's hours, with a person. If nobody can state the rule, it is not ready to automate.

**Should I build the whole automation plan at once?**

No. In the real run ChatGPT was asked for the smallest version and still proposed seven workflows. Build the one with no approvals and the lowest risk first, test it on a copy of your data, and add the next one once it works.

**Does ChatGPT build the workflow in n8n, Make or Zapier for me?**

Not in this lesson. ChatGPT writes the plan: triggers, steps, approval points and tests. You still build and test it in your own tool. Nothing in this lesson was built or run in n8n, Make or Zapier.

**What should I test before an automation goes live?**

Test the normal case and the cases that went wrong by hand, like a duplicate submission or a missing reply, on a copy of the spreadsheet with test people. The plan in this lesson also suggests one full cycle in shadow mode, where it works out who it would email but sends nothing.

[Start this module free](https://learn.ynteractive.com/modules/chatgpt-plan-automation-from-manual-process)

- Previous in ChatGPT for Everyday Work: [ChatGPT: Fix #N/A, #VALUE!, #DIV/0! and #REF! Errors in Excel](https://learn.ynteractive.com/content/chatgpt-fix-excel-formula-errors)
- Next in ChatGPT for Everyday Work: [ChatGPT: Write and Test the Prompt for an AI Step in n8n, Make or Zapier](https://learn.ynteractive.com/content/chatgpt-write-test-ai-step-prompt-n8n-make-zapier)

## More in ChatGPT for Everyday Work

- [Summarize a Meeting Transcript with ChatGPT](https://learn.ynteractive.com/content/chatgpt-summarize-meeting-transcript): Paste a long meeting transcript into ChatGPT and get the decisions, an action items table with owners and due dates, and a follow-up email ready to send.
- [Reply Faster: Turn a Messy Customer Email Thread into a Clear, Friendly Reply](https://learn.ynteractive.com/content/chatgpt-reply-to-customer-emails): Turn a messy customer email thread into a clear, friendly reply: give ChatGPT the thread and your facts, check every promise, fix the tone, and send it from Gmail.
- [How to Create a Project in ChatGPT for Your Brand Voice and Files](https://learn.ynteractive.com/content/chatgpt-projects-brand-voice): Set up a ChatGPT Project with your brand voice instructions, your price list and your style guide, so every new chat already knows your business.
- [Analyze a Spreadsheet with ChatGPT: Upload a CSV, Get a Chart and Three Takeaways](https://learn.ynteractive.com/content/chatgpt-analyze-spreadsheet-csv): Upload a CSV of sales to ChatGPT, have it check the data first, ask your questions, get a chart you can download, and finish with the three takeaways that matter.
- [ChatGPT Work: Have ChatGPT Do a Multi-Step Task](https://learn.ynteractive.com/content/chatgpt-work-multi-step-task): Switch ChatGPT to Work, hand it a real job (compare three suppliers into a spreadsheet), approve the sites it opens, skip a login, answer its question, check the file…
- [Get a Morning Invoice Summary in Slack using ChatGPT Dots](https://learn.ynteractive.com/content/chatgpt-dots-invoice-summary-slack): Set up a ChatGPT dot to check Gmail for new invoice emails and post one summary to Slack every weekday morning. A 7-step practice based on OpenAI's published screens.

[All 28 modules in ChatGPT for Everyday Work](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work)

## More ChatGPT tutorials

- [Get On-Brand Social Posts Every Time using a ChatGPT Custom GPT](https://learn.ynteractive.com/content/custom-gpt-social-media-assistant): See why generic prompts give generic results, then build a Custom GPT loaded with your brand voice and watch the output transform.
- [Prompt Writing 101: Role, Context, Task, Format](https://learn.ynteractive.com/content/prompt-writing-101-role-context-task-format): Ask ChatGPT for the same customer email twice, once vaguely and once with a four-part prompt, and see exactly why structure gets a better answer.
- [Send a Same-Day Follow-up After Every Discovery Call using ChatGPT](https://learn.ynteractive.com/content/chatgpt-discovery-call-follow-up-email-sales): Turn a discovery call transcript into the decisions, the next steps with owners and dates, deal notes for your CRM, and a follow-up email the prospect gets the same…
- [Hold Your Price When a Prospect Asks for a Discount using ChatGPT](https://learn.ynteractive.com/content/chatgpt-reply-to-discount-request-sales): A prospect wants 30% off and a free month to sign this week. Give ChatGPT the thread and your pricing rules, catch the discount you can't approve, and send a reply that…
- [Get Your Business Recommended by ChatGPT](https://learn.ynteractive.com/content/get-recommended-by-chatgpt): Ask ChatGPT who it recommends in your town, see which pages it trusts, then use it to write an FAQ page and Business Profile copy that answer real customer questions.
- [Run a Weekly Ops Review Where Every Task Has an Owner using ChatGPT](https://learn.ynteractive.com/content/chatgpt-weekly-operations-meeting-recap): Turn a weekly operations and vendor review into the decisions, the next steps with owners and dates, the risks to watch and a team recap, without turning a vendor's…

## Terms in this lesson

Plain-English definitions, each with an example from a free lesson.

[Zap (Zapier)](https://learn.ynteractive.com/content/glossary#zapier-zap) · [Automation](https://learn.ynteractive.com/content/glossary/automation) · [Email thread](https://learn.ynteractive.com/content/glossary#email-thread) · [AI agent](https://learn.ynteractive.com/content/glossary/ai-agent) · [Chatbot](https://learn.ynteractive.com/content/glossary) · [Temporary Chat](https://learn.ynteractive.com/content/glossary#temporary-chat)

[The full AI glossary](https://learn.ynteractive.com/content/glossary)

## Browse related tutorials

[ChatGPT for Everyday Work (28)](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work) · [ChatGPT tutorials (74)](https://learn.ynteractive.com/courses/chatgpt-for-everyday-work)
