Every team has a list of things they'd automate "if there were time." The problem isn't finding candidates — it's that most of the obvious ones aren't actually worth it, and the ones that are worth it are rarely the ones that get mentioned first.
Three questions before anything gets automated
Before we build anything, every task on the list gets run through the same three questions. It's not complicated, but skipping it is how teams end up automating the wrong things.
1. How often does this actually happen?
A task that happens twice a year isn't worth a build, no matter how painful it is in the moment. Automation pays off on frequency — daily or weekly tasks are where the time savings compound fast enough to justify the setup cost.
2. How much judgment does it require?
Tasks with a clear, repeatable structure (pulling numbers into a report, routing a lead based on fixed criteria) automate cleanly. Tasks that require reading between the lines — deciding whether a piece of content matches a client's mood that week — don't, at least not without a human checking the output.
3. What does it actually cost when it's late or wrong?
A delayed internal report is an annoyance. A delayed client-facing report is a relationship problem. The cost of the task going wrong should shape how much review stays in the loop, not just whether it gets automated at all.
What this looks like in practice
Running a real team's task list through these questions usually surfaces two or three things worth automating first, and they're rarely the ones anyone complained about loudest. Reporting is a common one — high frequency, low judgment, painful when late — and it's almost always underestimated as an automation candidate because it doesn't feel like "real" automation the way a flashy AI agent does.
The workflow audit isn't glamorous. It's the difference between automating things that save real hours every week and automating things that were fun to demo once.