Automating business processes
Automate the process, not the mess
Automating a broken workflow just makes it faster and bigger. Fix the process first, then hand it to AI.
· 4 min read · Process automation · Operational efficiency · Business process improvement
You have a process that takes someone 6 hours a week. You think: if I automate this, I save 6 hours a week. That math works only if the process is already good.
If it's broken, automating it just makes the broken thing faster. And more expensive to fix later.
Here's what usually happens: A property manager in Surrey takes referrals from three sources—email, a web form, a text from a contractor friend. She copies each one into a spreadsheet, adds it to a calendar, sends a confirmation email, and flags it for follow-up. That's 15 to 20 minutes per referral, and she gets 8 to 12 a week.
Someone suggests automation. "We'll pull the emails, read the form, scrape the texts, put it all in one place, send confirmations automatically." Sounds good. So she builds it.
Three months in, she's got a tool that now processes 40 referrals a week—because it's faster, so more come in. But the tool doesn't know which referrals are from contractors she trusts. It doesn't catch the ones missing a phone number. It doesn't know that texts from one builder always need a site visit first, while emails from another builder are always straightforward. The automation is fast but dumb. She's now drowning in 40 referrals a week instead of 12, and most of them need manual fixing before they're useful.
She automated the mess instead of cleaning it first.
The three questions
Before you automate anything, ask these:
Is this step necessary? A dental clinic in Kelowna prints every patient form, files it in a folder, then scans it back into the computer. That's two people, 4 hours a week. Automating the scanning saves 2 hours. Removing the printing and filing saves all 4. Ask why each step exists before you automate it. You might find that half the steps are there because "that's how we've always done it."
Can a person do it better? An accounting firm in Burnaby gets invoices from clients in different formats—PDFs, scans, photos, spreadsheets. They want a tool to read them all and file them automatically. But invoices from one client are always messy scans with handwritten notes. A tool will miss the notes. A person won't. Better to automate the clean invoices and have a person handle the messy ones. Automation isn't all or nothing.
What happens if it breaks? A logistics company in Calgary automates the dispatch of delivery routes based on address, time window, and vehicle size. If the tool gets the address wrong, a driver shows up at the wrong place. If it doesn't check whether a vehicle is actually available, it sends two routes to the same truck. Before you automate, know what breaks and how you'll catch it. Keep a person in the loop for anything that touches money, a customer, or a decision that matters.
How to fix the process first
Map the real workflow, not the one you think you have. Sit with the person who does the work. Watch them do it once. Write down every step, every decision, every tool they touch. This takes 30 to 60 minutes and saves weeks of wrong automation later.
Then ask: Which steps are waste? A plumber in Richmond spends 20 minutes a day checking three different apps to see which jobs are booked. Consolidating the apps removes that step. Which steps could be simpler? An accounting firm in Burnaby has a rule: "If an invoice is over 10 pages, print it and file it by hand." That rule exists because the scanner jams on thick documents. Fix the scanner, remove the rule.
Test the fix manually before you automate it. If you're removing a step, have the person skip it for a week and see what breaks. If you're changing the order of steps, do it by hand first. You'll find edge cases and gotchas that no one mentioned in the meeting. A property manager might discover that one builder always sends referrals without a site address, so you need a backup step to call and ask.
Once the process is clean and tested, automation becomes simple and reliable. The tool now does something straightforward. It breaks less often. It costs less to build. And it actually saves time.
Where to start
- Pick one process that eats 4 or more hours a week. Watch the person who does it for one full cycle. Write down every step and every tool. This is your real map.
- Ask: Which steps are necessary? Which ones could a person skip without losing anything important? Remove two or three steps, have the person do the process the new way for a week, and write down what breaks.
- Once the process is simpler and tested, write a one-page description of what you want to automate and what should stay manual. Share it with the person who does the work. If they say "that won't work because," listen. They know things you don't.