Your first automation sets the tone for every one after it. Pick well, and it quietly works for months, the team starts asking what else could be handed off, and the second one is easy. Pick badly, and it breaks on the third day, someone has to clean up after it, and "automation" becomes a word people sigh at.
The difference is rarely the tool. It's the choice of job. Here's a simple way to make that choice, what to leave to people for now, and how to start small enough that the first one sticks.
Write down a week of repeated work
For one week, keep a list of everything you or your team does more than once. Don't judge as you go. Write down what starts it, what you do, roughly how long it takes and how often it comes up. Include the small things: copying an order number from an email into a spreadsheet, telling a colleague a payment came in, saving an attachment to the shared drive.
The list is usually longer than you'd guess. The big, painful tasks are rarely the whole story. It's the dozens of small ones that eat the afternoon.
Score each task on four questions
- Does it happen often? Weekly at the very least. Multiply how often by how long: a four-minute task done 30 times a week is two hours a week, every week, before counting the interruptions.
- Can you write the rule down? If you can explain it to a new colleague in a few sentences, with no "it depends", a workflow can follow it.
- Does it hurt when it goes wrong, or when it's forgotten? A lead nobody calls back, an invoice paid late, a figure typed into the wrong cell. Jobs that are boring and matter are the ones people get wrong.
- Is it easy to check? You should be able to see at a glance whether it worked: a row appears, a message arrives, a file is in the folder.
Here's how a few typical tasks score:
| Task | Often | Clear rule | Hurts when missed | Easy to check |
|---|---|---|---|---|
| Add each new payment to the revenue sheet | Yes | Yes | Yes | Yes |
| Tell the team in chat when a big order comes in | Yes | Yes | Yes | Yes |
| Save invoice attachments to the shared drive | Yes | Yes | Sometimes | Yes |
| Reply to customer complaints | Yes | No | Yes | No |
| Write the quarterly update for investors | No | No | Yes | No |
| Choose a new supplier | No | No | Yes | No |
Start with a task that gets four yeses. If several do, pick the one people complain about most. A first workflow that removes a real annoyance wins the team over faster than one that saves a little more time on something nobody minded.
Look for the hidden "it depends"
Some tasks look like rules and aren't. "Forward the important emails to the team" sounds simple until you try to write down what important means. When you hit one of these, you have two choices. Narrow it until it is a rule: "forward emails from these five customers" is one line. Or keep that part with a person for now, and automate everything around it.
A useful test: if two people on your team would handle the same case differently, it isn't a rule yet. Agree on how it should work before you automate it, or you'll get a fast version of the disagreement.
Leave these to people, for now
- Anything said to a customer in your name. A workflow can write the draft and leave it in your inbox. A person reads it and presses send.
- Anything that moves money or can't be undone. Let the workflow gather everything a person needs to decide, and let the person decide.
- The exceptions. The odd invoice in a foreign currency, the customer who pays by check. Automate the common case, and have the workflow flag the rest.
- Judgment that needs history. Whether to chase a customer who has paid late twice but always paid. That knowledge is in someone's head, not in the data.
Start smaller than you think
The safest first version of almost any workflow is one that notices something and tells someone. It can't do damage, and it proves the trigger works. Then add one step at a time, and let each step prove itself before the next.
For invoices, that might look like this. Week one: post each new invoice email in the finance channel. Week two: also save the attachment to the invoices folder. Week three: add a row to the payables sheet with the sender and the date. By the end of the month the whole chore is handed off, and at no point did anyone have to trust a big change all at once.
Before you let it start on its own, run it by hand against real data and watch each step happen. You'll catch the column that's in the wrong place, the sender who uses two addresses, the email with no attachment. It's far cheaper to find those while you're watching than in three weeks' time.
Check in after a month
Did it run every time it should have? Did anyone have to fix anything after it? How much time did it give back, and did anyone notice? If the answers are good, pick the next task from your list. If not, find out why before you build anything else. A first workflow that people trust is worth more than five that people double-check.
How Lodol fits
Everything above works whatever tool you use. In Lodol, a workflow can start by itself when something happens in apps like Gmail, Google Sheets, Slack, Stripe, Airtable and Box, and the Integrations page lets you browse every app by category and connect the ones you use from the same page. Press Run to try a workflow by hand and watch each step as it happens, with results from apps like Gmail, Slack and Stripe summarized in plain words, and pause or stop it if something looks wrong. New to the idea? Start with What Is Workflow Automation? A Plain-English Guide.
Make the list, then build the first one.
Create a workspace and start with the task that scored four yeses.