Fix the process before you automate it

You've identified a process that eats 8 hours a week. Someone is copying data from one system to another, or chasing down missing information, or re-entering the same details three times because no one knows which version is current. You think automation could handle it. You're right—but only if the process is actually sound.

Here's what happens when you automate a broken workflow: the tool does exactly what you told it to do, faster and more reliably than a person ever could. But if the instructions are based on a process that was never designed well, you've now built a machine that reliably produces the wrong result, or creates new problems downstream, or orphans important decisions.

A dental clinic in Kelowna tried to automate patient intake forms. The clinic was collecting the same medical history from patients every time they came in—at the front desk, then again with the hygienist, then again with the dentist. The practice manager thought automation could skip the duplicate steps. Instead, they built a tool that sent forms to patients before every appointment. The tool worked perfectly. But the underlying problem wasn't the forms themselves—it was that no one had decided which version of the patient's history was the source of truth. The automation just made the confusion happen faster. They had to pause, agree on one intake process, designate one person to maintain the records, and then rebuild the automation around that decision.

The same pattern shows up everywhere. A property manager in Surrey was automating tenant communication. The workflow included a step where the manager checked if rent was late, then sent a notice if it was. Sounds straightforward. But the "check" involved looking at three different spreadsheets because the accounting system, the bank feeds, and the lease file didn't talk to each other. Automating that meant building a tool to check three places and guess which was right. The real fix was consolidating the data source first, so the automation had one clear answer.

A distributor in Calgary wanted to automate order entry. Their sales team was taking orders by email, text, and phone, then manually entering them into the system. Automation seemed obvious. But when they mapped the actual workflow, they found that 20 percent of orders arrived without a customer number, 15 percent had incomplete addresses, and 10 percent were missing delivery dates. The team was fixing these by hand and guessing based on context. You can't automate a guess. They had to standardize the order form first, train the sales team to use it, and get agreement on what to do when information was missing—then the automation worked.

So before you automate, do this:

Map the actual workflow. Sit with the person who does the work and watch them do it once, or ask them to walk you through it step by step. Write down every step, including the ones that feel too small to mention. Include the parts where they stop and wait, or ask someone else a question, or check something twice because they're not sure.

Find the decision points. These are the moments where a human has to choose what happens next. "If the invoice is over $5,000, it needs approval" is a decision point. "We check the customer's history to see if they've ordered before" is a decision point. These are the places where unclear rules cause delays. Write them down and make them explicit. Get agreement on the rule. Assign one person to own that decision—not to make it every time, but to know what the rule is and to update it when business changes.

Remove steps that don't matter. You'll often find work that exists only because it always has. A step where someone prints a report and scans it back in. A manual check that duplicates what the system already did. A field that's filled in but never read. Ask why each step is there. If the answer is "I'm not sure," that's a sign the step might not be necessary. Remove it before you automate it.

Get agreement on the clean version. Before you build anything, make sure the people who do the work, the people who depend on the output, and the person responsible for the decision all agree that the process as you've described it is the right way. This takes 30 minutes in a meeting, and it saves weeks of rework later.

Once the process is clean and agreed, automation becomes straightforward. The tool does the routine work. A person reviews the results and handles exceptions. Decisions stay with the person accountable for them. The time you save is real, because you're not automating away a problem—you're eliminating the repetition that was slowing down a process that actually works.

A plumber in Richmond automated their quote follow-up. Before automation, they had a process: send a quote, wait three days, call if they hadn't heard back, wait three more days, send a second email, then give up. They tried to automate the whole thing. It didn't work because the real problem was that quotes were being sent to the wrong email address half the time, or to a committee with no clear decision-maker. They fixed those first—verified email addresses, identified the decision-maker, got clarity on timeline. Then they automated the follow-up reminders. The tool now handles the timing and the second email. A person still makes the call, because that's where the relationship lives. The automation saved 2 hours a week by eliminating the admin, not by replacing the judgment.

Where to start

• Pick one process that takes more than 4 hours a week and involves repetitive data entry or checking. This week, spend 30 minutes with the person who does it and write down the steps exactly as they happen, including the wait times and the decision points.

• Identify the three biggest delays or unclear moments in that workflow. Ask the team why each one exists. If you can't get a clear answer, that's a candidate for removal.

• Schedule a 30-minute meeting with the person who does the work and the person who uses the output. Walk through the process you've documented and ask if it's the right way. Note any steps they want to change before you think about automation.