What should a business automate first?

The best first automation is rarely the most impressive one. It's the boring task someone does by hand every day that nobody would miss doing. Here's how to find it.

The four tests

Run each candidate task through four questions. The best first automations pass all four.

1. Frequency. Does it happen often, daily or many times a week? Automating something that happens once a quarter rarely pays back.
2. Repetition. Are the steps the same each time? If every instance is different, it isn't ready.
3. Rules. Could you write down how the decisions are made? “Enquiries from this area go to this branch” is a rule. “It depends on the customer” usually isn't, yet.
4. Pain. Does doing it by hand cause delays, mistakes or missed opportunities? If not, it can wait.

Good first candidates

• Getting enquiries into the CRM. From web forms, email or social DMs, without retyping.
• Routing work to the right person or location, using rules you already apply in your head.
• Follow-up reminders that trigger from what has actually happened, not from memory.
• Status updates to customers or colleagues when a job moves forward.
• Moving data between two tools that should already be talking. See integrations.
• Pulling details out of documents that arrive in a consistent format.

Simplify the process before you automate it

Automating a messy process just produces a faster mess. Take a common example, an enquiry that arrives by email:

Current process — seven manual steps

1. Enquiry arrives by email
2. Employee reads it
3. Copies the customer details
4. Checks which branch or person covers the enquiry
5. Forwards the information
6. Updates a spreadsheet
7. Sets a reminder

Simplified and automated

1. Enquiry arrives
2. Relevant information extracted
3. Correct person or location identified
4. Customer record created or updated
5. Correct team member notified
6. Follow-up triggered

The person is still involved, but only at the point where they add value: dealing with the customer. Everything before that happens on its own, and nothing else in the business software had to change.

Do you need AI for it?

Usually not, at least not for a first automation. Most of the work above is handled reliably by plain rules. AI earns its place in steps that involve reading or judging language: understanding what a free-text email is asking for, classifying an enquiry, or summarising a long thread. If a normal rule can do the job reliably, adding AI just adds cost and unpredictability. More on that on our AI and workflow automation page.

How to tell whether it's worth automating

A quick, honest calculation using your own numbers:

1. Minutes the task takes each time.
2. × How many times it happens per week.
3. × What that time costs per hour (wages, or the work it displaces).
4. × Weeks in the year.

That gives you the yearly cost of doing it by hand. Add a sensible allowance for the mistakes and missed follow-ups it causes. Then compare it with the one-off build fee plus any running costs, such as automation-platform or usage fees. If the manual cost is larger within a timeframe you're comfortable with, it's a good candidate.

Measure it afterwards the same way: how long the task takes now, and how often things get missed. That way you know whether it worked, rather than assuming.

What not to start with

Anything still changing week to week, anything that depends on relationships or high-stakes judgement, and anything rare. We cover those in what processes should not be automated.

Not sure which of your tasks passes the four tests? Tell us what's taking too long. Custom process automation is priced individually, once we've mapped the process.