Skip to main content
Automating business processes

Automating business processes

Automate the process, not the mess

Automating a broken workflow just makes the broken parts faster. Fix the work first, then hand it to AI.

· 5 min read · Process automation · Workflow design · Operations · Business efficiency

You have a process that takes 8 hours a week. Someone says AI can do it in 20 minutes. You think about the time back. You think about the cost. You say yes.

Then the automation goes live, and you find out the process was broken all along.

Maybe the data was always inconsistent, and now the tool chokes on it. Maybe three people were doing the same work because nobody knew who owned it, and now the tool does it once and two people are confused. Maybe the process has a step that only makes sense because of a workaround from 2019, and the tool doesn't know about the workaround.

Automation doesn't fix a broken process. It locks it in. It makes it faster and harder to see, which is worse.

Why broken processes break harder when automated

A person can see that a step is pointless and skip it. A person can notice that the input data is wrong and ask a question. A person can hold context across three separate tasks and know when to bend the rule.

A tool cannot. A tool does exactly what you told it to do.

So when you automate a messy workflow—one with unclear ownership, unnecessary steps, or decisions that live in someone's head—the tool amplifies the mess. It runs the bad process at machine speed. It produces bad outputs faster. It creates data that downstream work has to clean up. It makes the problem invisible until something breaks.

Then you spend three weeks untangling what the tool is doing wrong, when the real problem was the process itself.

What a broken process looks like

Here's a real example from a property manager in Surrey. Tenant applications came in three ways: email, a web form, and a phone call. The office manager logged them in a spreadsheet. Sometimes. A leasing agent also kept a list on her own. A background check company had their own portal. Nobody agreed on what "approved" meant. The process took 5 days and nobody knew where an application was.

When they first asked about automating it, the instinct was to build a tool that ingests all three channels, deduplicates, and runs background checks. Faster. Cleaner.

But the real problem wasn't the channels. It was that nobody had agreed on the process. So the tool would have been fast at the wrong thing.

Instead, they spent a week defining: one intake form, one decision point (yes/no/maybe with a reason), one person responsible for each stage, and when the background check happens. Then automation made sense. A tool that reads the form, files it, and flags incomplete applications took 20 minutes a week instead of 4 hours. And it worked because the process was clear.

Another example: an accounting firm in Burnaby was invoicing clients manually. The partner would write an invoice in Word, email it, paste the amount into a spreadsheet, and send a follow-up email when payment was late. Three separate tools, one person's head holding it together. They wanted to automate it.

But first they had to ask: Why Word? Why not the accounting software they already pay for? Why are we tracking invoices two places? Why does one person do this instead of the bookkeeper? It took three conversations. Once they answered those questions, automation wasn't even necessary—they just started using the tool they already had. The time fell from 6 hours a week to 1 hour a week. No new tool needed.

How to fix the process before you automate

Spend a week on this. Not a month. A week.

Step one: watch it happen. Sit with the person who does the work (or do it yourself). Write down every step. Write down where they wait for someone else, where they have to find information, where they make a decision. Don't fix anything yet. Just see it.

Step two: ask why. For each step, ask: does this create value, or does it exist because of something else? Can two steps become one? Does this decision belong to this person, or is it someone else's call? Write down the answers. You'll find steps that exist only because of a tool that's no longer in use, or decisions that should happen earlier.

Step three: redesign on paper. Write the process the way it should work. Fewer steps. Clear handoffs. One person owns each part. Decisions happen when you have the information. Run it by the people who do the work. Ask them if it's realistic. Change it if it isn't. Then agree on it together.

This is the hard part, and it's not technical. It's about getting three people in a room and saying "we're going to do this the same way from now on." Most businesses skip this because it feels slower than just building the automation.

It's not. A week of clarity saves weeks of fixing a tool that's automating the wrong thing.

Step four: then automate. Once the process is clear and everyone agrees, automation is straightforward. The tool does what you designed, not what you inherited.

When not to automate

Sometimes the answer is no.

If a process only happens 3 times a month, automation probably isn't worth the effort. If it requires judgment calls that change based on context, a tool might remove the judgment you need. If the process is about to change (new software, new team, new regulation), wait until it settles.

Be honest about the math. If something takes 2 hours a month, and you spend 40 hours designing and building the automation, you're not breaking even for 20 months. If the person doing it is leaving anyway, maybe the answer is to hire someone to do it, not to automate it.

The goal isn't to automate everything. It's to automate the things that matter, and to do it in a way that actually saves time.

Where to start

  • Pick one repetitive task that eats at least 3 hours a week. Write down every step, in order, exactly as it happens now. Don't change anything yet.
  • Ask the person who does the work: which of these steps could disappear, and why do they still exist. Write down their answers.
  • Sketch the process the way it should work. Show it to the people involved. Ask if it's realistic. Change it until everyone agrees. That's your baseline for what to automate.

Want the repetitive work handled?

We map the manual steps, connect the tools you already have, and automate what should be automatic, with a person still in the loop where it matters.