How to Choose the First Process to Automate in Your Business

Every business has a dozen tasks that could be automated, and picking the wrong one first sours the whole idea. Our short answer on what to automate first gives the quick version and this guide gives the full method. The team loses faith after a messy project that saved nobody any time. This guide gives you a scoring checklist so you can choose your first process on evidence instead of gut feel. You finish with a ranked shortlist and one clear winner.

Step 1: Collect candidates

Spend 20 minutes with the people who do the work. Ask each person: "What do you do every week that feels repetitive, and what do you do that involves copying information from one place to another?"

Write every answer on a list. Do not judge yet. Typical candidates in a small Australian business include:

• Entering enquiry details from email or a web form into a CRM or spreadsheet.
• Sending quotes and chasing them.
• Raising invoices in Xero or MYOB from a completed job.
• Sending appointment confirmations and reminders.
• Chasing overdue invoices.
• Collecting documents from new clients.
• Compiling a weekly report from several systems.

Aim for 8 to 15 items. If you have fewer, ask again with a specific lens, such as "what do you do on Monday mornings?"

Step 2: Score each candidate

Score every item from 1 to 5 on the six questions below. A 5 is always the good answer.

Question	1 means	5 means
Frequency: how often does it happen?	A few times a year	Many times a week
Time: how long does one run take by hand?	Under a minute	Ten minutes or more
Clarity: can someone write the steps down?	"It depends" every time	Same steps nearly every time
Data: is the input structured and in one place?	Scattered, handwritten, inconsistent	A form, a spreadsheet or a system field
Risk: what happens if it goes wrong?	Money lost or customer harmed	A minor annoyance, easily fixed
Ownership: is one person responsible?	Nobody, or five people	One clear owner

Add the scores. The maximum is 30. Anything below 15 is not a first project.

Be honest on clarity and risk. These two catch most failed projects. If the steps change depending on who is asking, the process needs a fix before it needs a tool.

Step 3: Apply the veto rules

Some candidates should be removed whatever their score. Strike out any process where:

1. A mistake would be expensive or embarrassing and there is no review step. Anything touching payments, tax or legal advice needs particular care.
2. The process depends on a person reading tone, such as handling a complaint.
3. The inputs live on paper or in someone's head.
4. The owner does not want it automated. A reluctant owner will undermine the project.

Vetoes protect you. A first project that causes harm sets the whole effort back. If you are unsure what a sensible starting point looks like, see good first AI project.

Step 4: Estimate the payoff

For your top three, work out a rough payoff using your own numbers:

Runs per week:                 ___
Minutes per run, by hand:      ___
Hours saved per week:          runs x minutes / 60 = ___
Hourly cost of the person:     $___ (use their loaded cost)
Value of saved time per year:  hours x cost x 52 = $___

Treat the answer as a ballpark. Time saved only has value if the person uses that time on something useful, so ask what they would do with the hours. Sometimes the payoff is not hours at all. Faster replies to enquiries or fewer missed follow-ups may matter more, and those are covered in our guide to measuring whether an AI change is working.

Step 5: Check the tools you already have

Before you plan anything new, look at what you own. Your CRM, accounting software, booking system and email platform all have some automation. Tools such as Zapier, Make and n8n connect them, and Zapier vs Make vs n8n compares the three. A basic automation is often a few steps: when this happens in one system, do that in another.

Where the task involves reading messy text, such as sorting enquiries by type, an AI model like ChatGPT or Claude can handle the reading step inside the workflow. Keep the AI to the part that needs it and let plain rules do the rest.

Step 6: Pick one and write the goal

Choose the highest score that survived the vetoes. Then write a one-sentence goal with a measure, such as "Every web enquiry is logged in the CRM with the right category within five minutes, without anyone typing it." This sentence becomes your test later.

Write down what you will not do in the first round. A first project that quietly grows to five processes will stall.

Step 7: Map before you build

Before anyone touches a tool, map the process as it runs today. That means the steps, the decisions, the exceptions and who does what. Our guide on mapping a process before automating it walks through it. Mapping often shows steps you can delete, which is cheaper than automating them. The business process mapping article covers the formal notations, and the glossary has a short entry on process mapping.

Copyable scoring sheet

Process: ______________________
Frequency (1-5):   __
Time per run (1-5): __
Clarity (1-5):     __
Data quality (1-5): __
Risk (1-5):        __
Ownership (1-5):   __
TOTAL (max 30):    __
Veto triggered?    Y / N  (reason: __________)
Owner:             __________
Goal sentence:     ______________________

What to do next

If you have a shortlist and want a second opinion on what to automate and how, workflow automation from 21Labs starts with a free discovery. We look at your processes, tell you honestly which ones are worth automating and which are not, and give you a fixed price before any work starts. Use the form below and we reply within 24 hours.