If we started Luda over from scratch, knowing everything we know now, you'd probably think we'd be automating things left and right from the jump. We wouldn't. The first thing we'd automate is nothing. We'd do every task by hand, and we'd keep doing it by hand until we knew it well enough to build the automation properly.
That can sound odd coming from an automation agency, but it's the order we'd follow every time. People who skip it usually end up building the wrong thing and losing a lot of time fixing it later.
You don't really know a task until you've done it
A task always seems simpler in your head than it turns out to be in practice.
Take sending an invoice after you finish a job. On paper it's four steps: fill in the template, add the amount, email it to the client, and mark it as sent. Do it by hand for a few weeks and the rest shows up. One client needs a PO number on every invoice. Another pays half upfront, so they get two invoices instead of one. A third wants it sent to their bookkeeper instead of to them.
If you had automated this on day one, you'd have built the four-step version. It would work for some clients and get the others wrong, and you might not find out until someone's payment was late.
Do it by hand and write down every step
For the first stretch, do the task yourself every time it comes up. While you do it, keep a simple log. A doc or a note on your phone is enough. Write down:
- 1Each step, in the order you actually did it
- 2Which tools or accounts you had to open
- 3Anything that was different from last time
- 4Roughly how long it took
The third item matters most. Whatever changes from one run to the next is exactly what an automation has to handle, and you'll only find those differences by doing the work yourself.
After enough runs, the log stops changing. At that point you know how the task really works, and the log doubles as your build plan, because it lists every step the automation needs to cover.
How to tell when it's ready to automate
A task is ready when it comes up often, the steps mostly stay the same, and you could write out the whole process without guessing. That last part is the first question on our Before You Build checklist, and doing the task by hand is how you get to a real yes on it.
Some tasks won't make the cut. A few will change too much from run to run, or only come up once a month, and doing them by hand will stay the better option. You only find that out by doing them, and it saves you from building something you didn't need.
Doing the work yourself also means you know what the right result looks like. That makes it much easier to catch an automation that runs without any errors but still gets something wrong, which is the hardest kind of failure to notice. We go into that in What to Do When Your Automation Breaks.
This is slower at the start. You'll spend time on work you know a tool could eventually do, and building the automation is a lot more fun than doing the same task for the fifteenth time. Still, a few extra weeks of manual work usually costs less than rebuilding an automation that was based on the wrong idea of how the task works.
Where to start
Pick one task you do every week. Do it by hand for the next few weeks and write down every step as you go. When the log stops changing, automate that one.