JOURNAL
Published on
WHICH PROCESSTO AUTOMATEFIRST
You can make this choice tonight, on paper, without hiring anybody.
THE ANSWER
Start automating with the dullest, most frequent process, not the biggest one. Look for the best ratio of pain to difficulty. Something that repeats every week. Something whose rules can be said out loud. Something that touches few systems and has an owner. The biggest problem is usually the least ready.
AI-generated image
IN SHORT
Start with the dull, frequent job, not the big one.
Multiply times per month by minutes per run.
Rules still under discussion: wait, don’t automate.
THE CRITERION
HIGH PAIN, LOW DIFFICULTY
The first process isn’t chosen for importance. It is chosen for ratio. On one side, the pain: how often it repeats, how much time it takes. On the other, the difficulty: how many systems, how many exceptions, how many habits. What you want has high pain and low difficulty. Usually it is the dullest thing you do. (A dull job is one whose rules are already settled.)
THE SIGNS
HOW TO SPOT THE FIRST ONE
Five signs you count rather than judge. Four out of five is enough.
It repeats on a fixed rhythm
Every day, every week, with every new client. Only frequency pays an investment back: 3 times a year stays manual.
The rules can be said out loud
Whoever runs it explains it in 5 minutes without saying “it depends”. The rules exist: I write them where they run alone.
The data already exists
It sits in a field, not in somebody’s head. A value pulled from a PDF sits a level above.
A mistake shows up quickly
If a wrong result is obvious within a day, you can start. If it stays invisible for 3 months, you need checks first.
It has an owner
One person who decides on exceptions without convening anybody. Without them you are missing whoever accepts the result.
DO IT YOURSELF
FIVE STEPS ON ONE SHEET
One sheet of paper and a week of attention. The fourth step removes the wrong candidates.
List what repeats
For one week, note every task somebody does the same way twice. The small things too: they are what fills the day.
Multiply frequency by minutes
For each entry: how many times a month it happens, how many minutes it takes. The product is your pain. Use your own numbers, not averages from elsewhere.
Count the systems and the exceptions
For the 3 entries with the highest pain, count the tools and the variants. 2 systems and 2 exceptions is a project; 6 and 12 is another trade.
Drop what isn’t ready
Remove rules still argued over, data that lives in somebody’s head, tools about to be replaced. They go back in the queue until the conditions exist.
Take whatever is left at the top
Among the survivors, pick the most pain and the least difficulty. Write in 3 lines what ought to happen by itself.
AI-generated image
WAIT
WHEN A PROCESS ISN’T READY
Automating an unstable process freezes it at its worst. Sort something else out first.
The rules change every month
If the criterion is still under discussion, the code that applies it gets rewritten every time. Settle the rule, then automate.
Nobody can say what happens in the exception
“What if the client doesn’t reply?” gets 3 different answers. That isn’t an exception: it is a decision nobody has taken.
The information lives in people’s heads
If a status is known by asking somebody, you need the field first. An automation reads a system, not a custom.
The underlying tool is about to change
Building on a platform replaced in 3 months means paying twice for the same connection. Automate after the switch.
THE BIG ONE
NOT THE BIGGEST PROBLEM FIRST
The biggest problem is almost always the least ready. Not by accident: it touches everybody and its exceptions have never been lined up. Starting there means 2 months of discussion instead of delivery. A first project with nothing visible to show gets no second one. A dull, frequent process, solved completely, shows people that work can change without breaking. After that the big problem becomes tractable.
The first process doesn’t have to impress anyone: it has to work on Monday morning.
QUESTIONS
- Where do I start automating?
- With the most repetitive process that touches the fewest systems. Note what repeats for a week and multiply frequency by minutes. Take what is left at the top.
- What if the most painful process is also the hardest?
- Then you split it. Nearly every large process holds a small, exception-free piece: a notification, a document to collect. Start there.
- How many processes at once?
- One. The second is designed better once you have watched people use the first.
- How do I know whether I chose well?
- If nobody asks to go back after 2 weeks, you chose well. If somebody keeps the old spreadsheet “just in case”, there was an unwritten exception.
Once you have your candidate and the 3 lines that describe it, tell me what happens today. I’ll say whether it is the right first one, including when the answer is no.