Owner's AI Masterclass3 min read

Turning repeat work into routines

Part 4 of the Owner's AI Masterclass. How to turn a task you do every week into a written routine, then a saved AI workflow, then, if it earns it, an automation.

Code Hawker team18 November 2026 · Facts checked 2 November 2026
Masterclass part 04: Turning repeat work into routines

In part 3 you learned to write instructions AI can follow. This lesson turns repeated tasks into routines. The aim is to stop solving the same problem from scratch every week, and to know which tasks are worth automating.

Think of it as a ladder with three steps. Each step saves time on its own. Only some tasks should climb all three.

Step 1: write the task down

Pick a task from your "AI can do it" column in part 1. Write it as numbered steps, the way you'd explain it to a new staff member:

Weekly payment follow-up

  1. Open the outstanding invoices sheet every Monday.
  2. For invoices 7–14 days overdue, send a polite reminder on WhatsApp.
  3. For 15–30 days, send a firmer reminder and note it in the sheet.
  4. For over 30 days, flag to the owner. No message.
  5. Mark "reminded on [date]" in column F.

Note three things for every task:

  • Inputs: what information is needed (the invoice sheet, customer name, amount, due date)
  • Output: what "done" looks like (messages sent, sheet updated)
  • Exceptions: when a person must step in (disputes, big amounts, long-standing clients)

If you can't write it down clearly, it's not ready for AI. Writing it down often reveals that the task has no fixed rules yet, and that's useful to know.

Step 2: a saved AI instruction, run by hand

Turn the AI part of the task into a saved instruction with blanks:

"Write a polite WhatsApp payment reminder from [business name] to [customer first name] for invoice [number] of ₹[amount], due on [date], now [days] days overdue. Tone: [friendly / firm]. Include UPI ID [id]. Under 50 words. Hindi and English."

Run it by hand every week for a month. Keep notes: did it need editing? Did any case not fit the rules? Adjust the instruction and the written steps.

Many tasks should stay at this step. A saved instruction that turns a 15-minute job into a 3-minute job, with a person in control, is already a good result. It also needs no setup cost, no new accounts and no risk of automated mistakes.

Step 3: automate, if it earns it

A task is ready for automation when:

  • it happens often: daily, or dozens of times a week,
  • it ran the same way for a month at step 2, with few exceptions,
  • the cost of a mistake is low, or mistakes are caught before they matter,
  • the inputs are already digital and in one place: a sheet, a form, a booking system.

Then the options are:

  • Simple automation tools that connect your apps, for example "when a new row appears in this sheet, send this WhatsApp template". See n8n, Zapier or Make.
  • An AI agent if the task needs reading free text or making small decisions, for example understanding a customer's reply and rescheduling. See when an agent beats plain automation.

Keep the exception rules from step 1. Automation should hand exceptions to a person, not guess.

Common mistakes

  • Automating before writing it down. You'll automate the confusion.
  • Skipping the month at step 2. That month is where the rules get fixed.
  • Automating rare tasks. Setting up automation for something you do twice a month rarely pays back.
  • No owner. Every routine and automation needs a named person who checks it.

Checklist for this lesson

  • Choose two repeat tasks from your part 1 list
  • Write each as numbered steps, with inputs, output and exceptions
  • Create a saved AI instruction with blanks for each
  • Run them by hand for four weeks and note every edit
  • Count how often each task happens per week
  • Mark which (if any) meet the four tests for automation

Next in the masterclass: Part 5, using the data you already have, safely. When a task is ready to automate, our workflow automation service builds it, with prices published.

Written by the Code Hawker team

We've designed and built 300+ websites, apps and campaigns since 2016, and now build AI agents and automation for growing businesses.

Free 30-minute consultation

Want help putting this into practice?

Free 30-minute call · Reply within 1 business day · See our privacy policy. Prefer chat? Message us on WhatsApp.