Fix the process before you automate it

You know the feeling. A task eats 4 or 5 hours of someone's week. It's repetitive, it's boring, and it feels like the perfect job for automation. So you automate it. Then, three weeks later, you realize the tool is creating chaos because the process itself was a mess.

This happens more than you'd think. A property manager in Surrey automates rent collection reminders, only to discover the system is sending notices to the wrong addresses because no one ever cleaned up the tenant database. An accounting firm in Burnaby builds a tool to file expense receipts, then finds it's sorting them into the wrong categories because the receipt names are inconsistent. A dental clinic in Kelowna sets up automated appointment confirmations, but half the phone numbers are outdated.

The problem is simple: automation is honest. It does exactly what you tell it to do. If the process is unclear, the automation will be unclear. If the data is messy, the automation will spread that mess faster.

What breaks when you automate a bad process

When you automate something that's already broken, three things usually happen.

First, you move the problem downstream. Instead of a person catching the error before it matters, the automation sends it deeper into your workflow. A quote goes to the wrong email. An invoice gets coded to the wrong account. A customer gets contacted twice. Now you have more work, not less, because you have to find and fix the mistakes the tool created.

Second, you lose the person who was fixing it on the fly. Right now, someone is probably doing the task and quietly correcting errors as they go. When you automate, that person is gone, and the errors are no longer caught. You don't notice until the problem is big.

Third, you build something you can't easily change. Once the tool is live, changing it takes time and money. If the process itself is wrong, you're now locked into the wrong way of doing things.

The fix is not to avoid automation. It's to automate the right process, not the broken one.

How to audit a process before automating it

Start with the people doing the work. They know what's wrong. They live with it every day.

Spend an hour with the person or team doing the task. Watch them do it once, from start to finish. Don't interrupt. Then ask three questions:

• Where do you have to stop and ask someone else what to do next?
• Where do you have to look in multiple places to find the information you need?
• What mistakes happen most often, and how do you catch them?

Write down the answers. Those are your problems.

A plumber in Richmond who automates job scheduling might discover that technicians are guessing at drive times because the job addresses are incomplete. A distributor in Calgary automating order entry might find that no one agrees on what "rush" means, so orders get prioritized differently every time. A 12-person software company in Gastown automating invoice sending might realize that some clients need a different format than others, and no one documented which is which.

These are the things that will break your automation. Fix them first.

Three things to fix before you automate

One: clarify the steps. Write down what happens in order. Not what should happen. What actually happens. Include the decisions: If X, then do Y. If not X, then do Z. If no one can agree on the steps, the process is not ready to automate.

Two: clean up the data. If you're automating something that touches customer names, addresses, phone numbers, or account codes, spend a day cleaning them up. Delete duplicates. Fix typos. Fill in blanks. A messy database will break your automation faster than a messy process will.

Three: decide who decides. Some steps need a person to make a judgment call. A quote that's 20% higher than usual. A customer who hasn't paid in 90 days. A job that's more complex than the standard template. Decide now who checks these things and when. Don't make the automation decide. That's where bad decisions hide.

This work takes 2 to 5 days for a typical process. It feels slow, but it saves you 4 to 6 weeks of frustration after the automation is built.

What a clean process looks like

A clean process is one where the same thing happens the same way every time. Not because everyone is rigid, but because everyone understands the rule and agrees with it.

When you automate a clean process, the tool just does what you would do anyway. It's boring. It's predictable. It works.

A property manager in Surrey with a clean rent collection process has a database where every tenant record is complete and current. She has a written rule: send reminders on the 25th if rent is not paid by the 24th. No exceptions, no judgment calls. When she automates it, the tool sends the right reminder to the right person on the right day. No chaos. No cleanup.

That's the goal.

Where to start

• This week, pick one repetitive task and watch someone do it. Sit with them for 30 minutes and write down the steps, the decisions, and the places where they have to stop and ask for help. That's your audit.

• Ask them the three questions: Where do you stop and ask for help. Where do you have to look in multiple places. What mistakes happen most often. Write the answers down. These are your problems to fix, not for automation to handle.

• Spend a day fixing the three things: clarify the steps, clean up the data, and decide who decides. You don't need permission or a big project. Just do it. Then you'll know whether this process is ready to automate.