Automate after you document, not before

Part 6 of 9. The order is fixed: understand, document, assign an owner, then automate.

Office Clones
October 20, 2026

Every failed automation story sounds the same. Someone bought the tool, connected the apps, and three months later nobody trusts the system, so they went back to doing it by hand — now with an extra subscription.

Automating a chaotic process just makes the chaos faster. That is not a warning about technology. It is a warning about order.

The order is not a suggestion

Understand what actually happens, including the informal phone-call step nobody wrote down.

Document it, so the version being automated is the version people do.

Assign an owner, because a system with no owner is a system nobody is accountable for.

Then automate.

Skip step two and you are encoding a guess. Skip step three and you have automated the invisibility of a problem. Undocumented automation is how a business acquires a workflow that nobody can debug and everybody depends on.

Automate the repetitive

Good candidates share a shape: high frequency, low variation, clear success condition.

• Lead capture — every enquiry lands in one place with an owner attached
• Reminders — the follow-up happens because it is scheduled, not remembered
• Invoice follow-up — the sequence runs whether or not anyone is having a good week
• Onboarding steps — the same welcome, the same documents, the same first-week sequence
• Status updates — the customer knows where things stand without asking
• Reporting — the numbers arrive rather than being assembled

Bad candidates are the inverse: low frequency, high variation, judgment at every step. Automating those produces brittle rules that fail exactly when the situation matters.

Centralise the knowledge

Procedures, client instructions, pricing rules, brand guidelines, FAQs, responsibilities and decision logs belong in one searchable place.

Not six inboxes and someone's memory. Not a shared drive with eleven folders called "final."

Done properly, a knowledgebase stops being documentation and becomes how the business actually operates. The difference is whether people open it on Tuesday, or whether it is written for an audit that never comes. The test is the same as the last two parts: does a person who is not you find the answer without asking you.

Track the right data

Eight numbers are enough to run most businesses your size:

• Leads
• Conversion
• Revenue
• Profit
• Retention
• Completion time
• Overdue tasks
• Response time

More dashboards do not make better decisions. The right numbers do. A dashboard is a display; a decision is what happens after you look at one.

If you are going to track only two for the next month, take response time and overdue tasks. Both are leading indicators, both are almost always owner-caused, and both improve within a week of being watched.

The tool question

Understand the workflow before you choose the tool, not the other way around.

A tool solves a shape of problem. If you have not described the shape, every tool looks approximately right, which is why business software shopping feels like standing in a very expensive fog.

One more thing worth saying plainly: the tool is the cheap part. The order of operations is the expensive part, and no subscription includes it.

Do this today

Write down the one process you have automated, or tried to, that still needs a person to babysit it.

Then answer: which of the four steps did you skip? Almost always it is "document" or "owner."

That is the repair, and it is usually an hour of writing rather than a new tool.

What this is part of

Part six of nine in How to Be a Real CEO in 30 Days or Less. The Workflow Improvement Plan worksheet in the full guide asks for the trigger, tools, definition of done and documentation status — the four things that decide whether an automation survives.

[Get the full guide and the four worksheets →]

Next Thursday: delegate outcomes, not tasks. It is a different instruction, and it produces a different result.