BeginnerLesson 2 of 305 min read

Mapping a Process Before You Automate It

Map a week of work as a table, flag rule-based steps and let a short program find your best first candidate for automation.

In this lesson you will

  • Write a process as timed steps
  • Flag steps that follow fixed rules
  • Find the safest, biggest first candidate
  • Know which steps to leave to people

About 20 minutes

The mistake of automating a mess

Ravi is excited. He wants to automate everything in the shop by Friday. His friend Mira, who works in operations, asks him a question first: ‘Can you write down every step you do, in order, on one page?’ Ravi tries and discovers that he cannot. Some steps live only in his head, others change depending on the customer, and he has never timed any of them. Mira’s point is a famous one in automation: if you automate a confused process, you simply get confusion at machine speed. So before touching any tool, you map the process.

Write the process as a list of steps

A process map does not need special software. A table on paper works. For each step, note four things: what you do, how long it takes each time, how many times a week it happens and whether the step follows fixed rules or needs your judgement. That last column is the one that decides what can be automated.

  1. List every step from the moment a job starts to the moment it ends.
  2. Time each step honestly, using a clock for a few days if you can.
  3. Count how many times it happens in a normal week.
  4. Mark whether it follows fixed rules, needs judgement, or both.
  5. Total the minutes, then see which steps give the biggest safe saving.

Let a small program do the sums

Here are six made-up weekly tasks for the print shop. Each row holds the minutes per time, the times per week, and two yes or no flags: does it follow fixed rules, and does it need judgement. A task is a candidate for full automation only if it follows rules and needs no judgement. Tasks that need both are better as partial automations, where the machine does the first draft and a person decides.

# MADE-UP weekly steps: (step, minutes each time, times a week, follows fixed rules, needs judgement)
steps = [
    ("read and sort emails", 2, 25, True, False),
    ("reply to simple quote requests", 6, 10, True, True),
    ("check payment status", 4, 5, True, False),
    ("decide special discounts", 8, 2, False, True),
    ("check design proofs", 30, 6, False, True),
    ("copy order details to the job list", 3, 15, True, False),
]
total = sum(m * t for _, m, t, _, _ in steps)
easy = [(n, m * t) for n, m, t, rule, judge in steps if rule and not judge]
easy_total = sum(x for _, x in easy)
print("minutes a week in all:", total)
print("fully rule-based minutes:", easy_total)
print("share:", f"{easy_total / total:.0%}")
best = max(easy, key=lambda x: x[1])
print("best first candidate:", best[0], f"({best[1]} min)")
print("hours a month, rule-based:", round(easy_total * 52 / 12 / 60, 1))
mixed = [n for n, m, t, rule, judge in steps if rule and judge]
print("needs a human check:", mixed)
minutes a week in all: 371
fully rule-based minutes: 115
share: 31%
best first candidate: read and sort emails (50 min)
hours a month, rule-based: 8.3
needs a human check: ['reply to simple quote requests']

The shop spends 371 minutes a week on these tasks. Only 115 of them, which is 31% of the total, are fully rule-based. The biggest of those is read and sort emails at 50 minutes a week, so that is the best place to start. Across a month the rule-based work is about 8.3 hours, a real chunk of time, though not the whole job. The program also flags the steps that need a human check: reply to simple quote requests. Replying to quote requests follows rules, since the questions repeat, but a person should still approve the price. That makes it a candidate for a draft written by the machine and checked by Ravi.

Notice the biggest time sink of all, checking design proofs at 180 minutes a week. It needs judgement and does not follow fixed rules, so no automation here should touch it. Mapping tells you honestly what to leave alone, and that is as valuable as what to automate.

Questions to ask about every step

  • What starts this step? Is the trigger clear and reliable?
  • What information does the step need, and where does it come from?
  • What does the step produce, and who or what uses it next?
  • What happens if the step goes wrong, and how would we notice?
  • Would a stranger be able to follow the written steps?

If a step fails the last question, tidy the process before automating it. Often this alone saves time, with no tool at all.

Common mistake: automating what annoys you

Judging by feeling instead of by time. The task that annoys you most is often not the task that costs the most minutes. Measure first, then choose.

Made-up sample data

All tasks, minutes and counts here are made up. Your own shop, office or studio will look different. Remember that building and maintaining an automation also takes time, so count that cost too.

Draw your own six-row table for one week of work. Add the flags. Then circle the one rule-based step with the largest total minutes. That is your first candidate.

Practice for this lesson

Check what you learned

Five quick questions. Pick an answer and you will see straight away whether it is right and why.

Why map a process before automating it?

  1. Tools refuse to start without a map
  2. Automating a confused process just makes confusion faster
  3. It makes the tool cheaper
  4. It removes the need for testing

You must know the steps, timings and rules before a machine can follow them.

Which flags make a step a candidate for full automation?

  1. Needs judgement only
  2. Takes a long time
  3. Follows fixed rules and needs no judgement
  4. Is disliked by staff

Fixed rules without judgement are what a machine handles well.

What is a good way to automate a task that needs both rules and judgement?

  1. Machine drafts, a person decides
  2. Automate it fully
  3. Ignore it
  4. Do it twice

A draft saves time while a person keeps control of the decision.

In the sample, which rule-based task cost the most minutes a week?

  1. Checking design proofs
  2. Deciding special discounts
  3. Reading and sorting emails
  4. Checking payments

It came to 50 minutes a week, the largest of the rule-based tasks.

Why leave design proof checking alone?

  1. It needs judgement and does not follow fixed rules
  2. It is too short
  3. It is illegal to automate
  4. It happens too rarely

Taste and judgement stay with people.

Summary

  • Map and time the process before choosing a tool.
  • Fixed rules without judgement suit full automation.
  • Rules plus judgement suit machine drafts and human decisions.
  • Count the cost of building and maintaining too.

Next lesson

Lesson 3 looks at what no-code platforms actually do and how steps depend on each other.

Questions

How long should I time my tasks for?

A few normal days is enough to start. The goal is a fair estimate, not perfect precision.

What if my process is different every time?

Then it is probably not ready to automate. Look for a smaller, repeatable part first, or write down the rules you secretly use.

Are the minutes in the lesson real?

No. They are made-up sample figures for a pretend print shop.

Progress is saved in this browser.