Automation: where you actually start

When a company decides to automate something, instinct points at the most important process: the one that brings in the money, the one that is a disaster if it jams.

That is precisely the one not to start with.

The three conditions

A process is a good first candidate when it has all three of these.

It is repetitive

It happens many times, always the same way. If it happens three times a year, automating it costs more than it saves — even when it works perfectly.

It is checkable

You can look at the result and say immediately whether it is right or wrong. A draft email is read in ten seconds. A sales forecast is not: you find out in six months, and in the meantime nobody knows whether the system is working.

This is the condition that matters most. An automation that gets things wrong without anyone noticing is worse than doing the work by hand.

The cost of a mistake is low

If it gets something wrong, somebody notices before it becomes a problem. A bad draft is rewritten; a bad payment is chased.

Why not the most important process

For two reasons.

The first is technical: on a critical process you cannot afford mistakes, and a first automation project always gets something wrong — not through incompetence, but because the exceptions only surface in use.

The second is organisational, and it matters more. The first automation exists to show that this works. If the first attempt touches the critical process and causes a problem, there will be no second attempt: the belief that “these things do not work” has formed inside the company, and undoing that costs more than building the system.

Where to start, in practice

In most of the companies we work with, the first process is one of these:

• The first reply to an enquiry. The email arrives, the system works out which job it is about, and drafts the reply with the right quote already attached. A person confirms the draft.
• Filing what comes in. Documents and photos landing in the right folder under the right name, instead of wherever.
• Data from one form to another. The copy-and-paste between two programs that do not talk: the most mindless thing a person can do, and the easiest to remove.

The measurement to take first

Before automating anything, count how many hours a week you spend on it. Not by eye: for one week, write it down.

That number is the only thing that says whether the automation is worth its cost, and in half the cases it says no. That is a good result: you have saved yourself the project.