Automation should remove work, not add software

Ask a struggling operation what tools it runs and you will usually get a list of eleven. A CRM, a scheduler, a form builder, a helpdesk, two chat apps, a project tracker, an invoicing tool, a spreadsheet that quietly runs the actual business, and two subscriptions nobody can account for.

Every one of them was bought to save time. Collectively they cost more time than they save, because the work has moved. It is not in the tasks any more. It is in the gaps between the tools, and the gaps are staffed by people copying things.

Copy-paste is the tell

The fastest audit I know: for one week, have everyone note every time they move information from one system to another by hand. Not fix it, just note it.

The list will be longer than anyone expects and it will be boring, which is exactly why nobody has looked at it. Each individual instance takes ninety seconds and feels too small to raise. Multiply it by five people and two hundred and fifty working days and it stops being small.

That list is your automation backlog. It is more useful than any vendor discovery call, because it came from the actual work.

Subtract before you add

Before automating a task, try to delete it. A surprising share of recurring work exists because someone asked for a report in 2023 and nobody has asked whether it is still read.

Automating something nobody needed is worse than leaving it manual, because manual work eventually gets questioned. Automated work runs silently forever.

The order I use:

• Delete it. Is anyone actually consuming this output?
• Simplify it. Can the process have fewer steps or fewer approvals?
• Consolidate it. Can two tools become one, so the gap disappears rather than getting bridged?
• Then automate what survives.

The measure that matters

The right question after an automation project is not how many workflows are running. It is whether anyone got their week back.

If headcount is the same, hours are the same, and the only change is that there is now a twelfth tool with a dashboard, the project did not work. It just moved the cost from labour to licences.

Automation that succeeds is boring to look at. Something that used to require a person no longer does, and there is nothing new to check.