Most business owners come to automation with the same question: "What should I automate?" It's the wrong first question. The better one is: "Which of my tasks are repetitive and low-judgment, and which ones need a human brain making a real decision?"
Get this sorting wrong and you end up in one of two bad places. Either you automate something that needed a human touch — and your customers notice — or you keep a person manually doing something a system could handle in seconds, burning hours you're paying for.
The test: repetition vs. judgment
A simple way to sort any task is to ask two questions:
- Does this task follow the same steps almost every time?
- Does getting it right depend on context only a person would know?
If the answer to #1 is yes and #2 is no, it's a strong automation candidate. If #2 is yes — even if the task looks routine — a human probably needs to stay in the loop, at least for now.
Good first candidates for automation
These are usually safe places to start, because the cost of a small mistake is low and the steps rarely change:
- Appointment reminders and confirmations. Nobody needs judgment to send "see you tomorrow at 2pm."
- Lead intake and routing. A form fills out, the right person gets notified, the lead lands in your CRM. No decision-making required — just moving information correctly.
- Invoice generation and payment follow-ups. The math and the timing are fixed; automating this frees up real hours.
- Data entry between tools. If you're copying information from one system to another by hand, that's hours you're not getting back.
- Basic customer FAQs. Hours, pricing tiers, return policy — questions with one correct answer, asked over and over.
What should usually stay human
- Anything involving a complaint or an upset customer. Templates can help you respond faster, but the read of tone and the decision on how far to go to fix things needs a person.
- Pricing or contract exceptions. The moment a client asks for something outside your standard terms, that's a judgment call, not a workflow.
- Final review of anything public-facing. Automated drafts are fine; automated publishing without a human check is where mistakes become visible.
- Sensitive HR or client relationship conversations. These carry weight a script can't carry.
The middle ground: automate the draft, not the decision
A lot of tasks don't have to be all-or-nothing. You can automate the first 80% — pulling data, drafting a reply, flagging an issue — and still have a person make the final call. This is often the most realistic starting point for small teams: it removes the grunt work without removing accountability.
A practical next step
Pick one week and write down every recurring task your team does. Sort each one using the two questions above. You'll likely find that 3-5 tasks are ready to automate now, a few more could move to the "draft, then human reviews" middle ground, and the rest genuinely need a person.
Start with the shortest, clearest task on your automation list — not the biggest one. A small, working automation you trust is worth more than an ambitious one that sits half-finished. If you want a second pair of eyes on that sort, that's exactly the kind of conversation worth having before you build anything.