Automate the right process, not the broken one

When you automate a bad process, you don't fix it. You make it worse. You make it faster at doing the wrong thing, and now the wrong thing happens at scale.

A property manager in Surrey started automating her tenant inquiry emails. The system was fast: it sorted, filed, and routed inquiries to the right team member in seconds. But it routed them to the wrong person half the time because the original process had no clear rule for who handled what. The automation just made the routing faster and the confusion louder.

An accounting firm in Burnaby tried to automate invoice entry into their billing system. The system read the invoices perfectly. But it couldn't decide which cost code to use because the firm had never written down that rule—each bookkeeper used their own judgment. The automation stopped and asked for help 40 percent of the time. They'd spent money to make the work slower.

These aren't failures of the automation. They're failures of the process.

What breaks when you automate a mess

Automation is honest. It does exactly what you tell it to do. If your process has hidden rules, fuzzy handoffs, or decisions that live in someone's head, automation will hit those walls and stop. Or worse, it will push through and create chaos downstream.

The most common break points are:

• Missing decision rules. Who decides if this invoice gets paid today or next month. Who decides if a client gets the standard quote or a custom one. If the rule isn't written down, automation can't follow it.
• Handoffs without clarity. Work moves from one person to the next, but nobody knows when. The second person doesn't know when to start. The first person doesn't know when to hand off. Automation gets stuck in the gap.
• Steps that aren't steps. Someone reviews something "to make sure it looks right." That's not a step. That's a feeling. Automation needs to know what right looks like.
• Data that lives in different places. One team member keeps notes in email. Another uses a spreadsheet. A third uses the CRM. Automation can't read three languages at once.

When you try to automate over these problems, the automation either fails (and you've wasted time and money), or it succeeds and creates new problems (wrong decisions made fast, work routed to the wrong place, data lost in translation).

How to find the real bottleneck

The bottleneck you see is often not the real one.

You think the bottleneck is the data entry. A 12-person software company in Gastown was spending 6 hours a week entering timesheets into their invoicing system. That seemed like the obvious place to automate. But when we watched the process, the real bottleneck was 90 minutes of back-and-forth emails asking people for missing information. The data entry was fast. The waiting was slow.

You think the bottleneck is speed. A plumber in Richmond wanted to automate job scheduling because dispatch was taking too long. But the real bottleneck was that the scheduler didn't have enough information to make good decisions—no data on which crew was closest, which jobs were urgent, or which crew had done similar work before. Speed wouldn't help. Clarity would.

You think the bottleneck is volume. A dental clinic in Kelowna wanted to automate appointment reminders because sending them manually took time. But the real bottleneck was no-shows, not the sending. Automation could send reminders faster, but it wouldn't solve the no-show problem unless the reminders actually changed behaviour (which they might not).

To find the real bottleneck, don't ask. Watch.

Spend an hour watching the process happen. Or ask someone to walk you through one real example, step by step, with all the waiting and deciding and checking included. Write down where time actually goes. Write down where someone has to make a choice. Write down where information gets lost or duplicated. That's where your real bottleneck is.

How to fix the process before you automate

Once you see the real bottleneck, fix it first.

Write down the decision rules. If someone has to choose, write the rule. If the rule is "use your judgment," write down what good judgment looks like. Give examples. A distributor in Calgary had a process where the warehouse team decided which orders to pick first. The rule was "pick urgent orders first." But urgent meant different things to different people. They wrote it down: orders marked urgent by the customer, or orders for same-day delivery, or orders over 50 units. Now everyone picks the same orders first.

Remove steps that don't add value. In the Surrey property manager's process, every inquiry got logged, then sorted, then routed, then logged again. The second logging didn't add value. They removed it. Now the process is shorter and clearer.

Clarify the handoff. Who hands off to whom, and when. What does the next person need to know. Write it down. A Burnaby accounting firm had invoices bouncing between people because nobody knew whose job it was to match them to purchase orders. They assigned it: the first person to touch the invoice does the match. Now it happens once.

Centralize the data. If information lives in three places, move it to one. This doesn't have to be complicated. It can be a shared spreadsheet. It can be a form that feeds into one place. But it has to be one place.

Once you've done this, the process is clearer, shorter, and easier to automate. The automation will work because it has clear rules to follow.

Where to start

• Pick one process that wastes your team's time. Not the biggest one. One that takes 3 to 6 hours a week and involves clear, repetitive steps. Write down what happens now, step by step, with one real example.
• Spend one hour watching or walking through it. Ask the person who does it to show you one complete example. Note where they wait, where they decide, where they check something. That's where the real work is.
• List the three biggest bottlenecks. For each one, write one sentence about why it happens. Is it a missing rule, a unclear handoff, a fuzzy decision, or data in the wrong place. That's your fix list.