Automation is not a strategy. It is a multiplier — and multiplying a bad process just gets you more of it. These are the jobs where it consistently wins.
Worth automating
- Lead intake. Form to CRM to tagged, assigned and acknowledged, in seconds rather than the next time someone checks the inbox.
- Appointment logistics. Booking, reminders, reschedules, no-show follow-ups. Pure admin, zero judgement required.
- First-draft writing. Proposals, follow-up emails, product descriptions — drafted by AI, edited by a human before it ever goes out.
- Recurring reports. Pull the numbers, format them, send them on a schedule. Nobody should be copying cells on a Friday afternoon.
- Data clean-up. Deduplicating, normalising, enriching records that would otherwise rot.
Not worth automating yet
- Anything a customer would notice was automated and mind — complaints above all.
- Judgement calls: pricing exceptions, hiring, anything with legal weight.
- Processes you do twice a year. The set-up costs more than the task.
- Steps nobody has written down. Document first, then automate.
A simple test
Multiply how long a task takes by how often you do it. If it clears two hours a month and the rules fit on an index card, automate it. If not, leave it alone and go find a bigger one.
Keep a human at the end
The best set-ups are not hands-off, they are hands-light: AI does the fetching, drafting and sorting, a person makes the call. That is where the time savings are real and the mistakes stay small.