Which business processes should you automate first?
Short answer. Automate first the tasks that are repeated every week, take real hours, cause costly mistakes when done by hand and follow clear rules. Re-keying between systems, building invoices from completed jobs, sending reminders, assembling reports and generating documents usually top the list. Keep judgement calls, exceptions, customer relationships and large approvals with people. List the tasks, score them, start with one, and measure the hours before and after.

In this guide · 5 minute read
Start with a list of what people repeat
Ask each team to write down the tasks they do every day or every week that feel like copying, chasing or checking. Don't filter yet. The list from an accounts team might include typing supplier invoices into Xero, emailing overdue reminders and building the weekly debtors report. Operations might add retyping booking emails into the job system and sending confirmation texts.
For each task, note who does it, how often, roughly how long it takes, and what goes wrong when it's done badly. Fifteen minutes with each team usually produces a longer list than anyone expected, and it's the raw material for every decision that follows.

Score each task on three things
Give each task a rough score, out of five is fine, on three questions. The aim is a ranking you can argue about around a table, and precision doesn't matter much. Have the people who do the task score it, because they know where the time and the mistakes really go.
- 01Hours per week: how much time it takes across everyone who does it.
- 02Cost of an error: what a mistake costs in money, rework, customer goodwill or compliance.
- 03How rule-based it is: whether you could write the steps down so a new starter follows them without asking.
Good first candidates
Tasks that score well on all three tend to be the same across industries, whether you run a cleaning company, a freight business or an accounting practice. They're tedious for people, the rules are already known, and the result is easy to check.
- 01Re-keying between systems, such as jobs into accounting or orders into a courier portal.
- 02Creating invoices from completed jobs, timesheets or delivery dockets.
- 03Reminders and chasing: overdue invoices, unsigned quotes, expiring licences, missing timesheets.
- 04Assembling the weekly or monthly report from several sources.
- 05Generating documents such as quotes, contracts, compliance certificates and job packs from data you already hold.
- 06Triage of incoming enquiries: reading the email or form, working out what it is, and putting it in the right queue with the details filled in.
★★★★★
“One of the BEST companies our business has worked with.”
What to keep with people
Automation is good at doing the same thing correctly every time. It's poor at noticing that this time is different. Keep people on decisions that need judgement, such as whether to take on a risky job or extend credit to a struggling customer. Keep them on exceptions, where the rules don't fit. Keep them in customer relationships, where a phone call from someone who knows the account matters. And keep approvals above a threshold with a named person, so a large credit or price change always gets a second look.

Rules-based automation or an AI agent?
If the input is already structured and the steps never change, plain rules-based automation is the right tool. When a job is marked complete, create the invoice. When an invoice is fourteen days overdue, send the reminder. It's predictable, cheap to run and easy to test.
AI agents suit work where the input arrives messy: a booking buried in an email thread, a supplier invoice as a PDF, a technician's notes and photos. The agent reads it and drafts the record, and a person approves anything that matters. Inside your own system, chosen decisions wait for a person and every action the agent takes is logged.
Measure before and after
Before you automate anything, time the task for a week or two and count the errors. Without that baseline, nobody can say later whether it worked. After the automation goes live, measure the same things: hours spent, mistakes caught, how long the task now takes from start to finish. Check what the freed-up time is being used for, too. If the hours saved disappear into new work nobody planned, that's worth knowing before you choose the next process to automate.
Terms in this guide