Which jobs are suited to automation, and which are not?

When we ask a company “which of your jobs keep repeating?”, the list usually turns out long: preparing quotes, emailing suppliers, entering a list into a system, the monthly report, pulling information out of documents. But not every job on the list is equally suited to automation. Some become a tool in a few days; others first need the work itself to be sorted out.

This article is meant to help you look at your own work and answer the question “could this become a tool?”. We cover the choice between an off-the-shelf package and a tool built for your company separately in Off-the-shelf software or custom-built?

Three tests: repetition, rules, digital data

1. Does the job repeat often?

Having a tool built for a job that is done once a year and takes half a day usually does not pay for itself. A job done every day or every week, going through the same steps each time, is the first candidate. The measure is simple: how many hours in total does this job take each month? The more hours, the more the tool is worth.

2. Can its rules be written down?

If the person doing the job can write the steps down, the job runs on rules: “If the value in this column is over 100, add it to this list; otherwise, to that one.” If the steps are often interrupted by “then I look at the customer and decide depending on the situation”, the job is full of decisions. The part that makes decisions stays with people; the part that can be tied to a rule moves to the tool.

3. Is the data already digital?

A tool works with the data it can read. If the information sits in Excel, an email, a PDF or a system, the job gets easier. If it sits in paper folders or in one person's memory, the data has to be digitised first. Even scanned documents are a step ahead in this respect; with AI-assisted reading, hundreds of documents can be turned into a single table. We explain step by step how this works on the Documents to spreadsheet page.

Examples of jobs that suit automation

• Entering a list into a system. Typing a list of thousands of rows, one by one, into an official system or an in-house program. The rules are clear, the data is digital, the repetition is high. Bulk data entry is a typical example of this kind of work.
• Cleaning up duplicate records. Finding five different spellings of the same customer. The matching rule can be written down; the decision to delete is left to a person. See Record cleanup.
• Producing quotes and reports. You choose the customer and the product, and the document comes out in the same layout every time.
• Reminder emails. A personalised reminder at set intervals to a supplier who has not replied.

Jobs that do not suit automation, or need preparation first

• Jobs that need a different decision every time. Giving a particular customer a special discount, deciding how to resolve a complaint. Here a tool can gather the information and put it in front of you; it cannot make the decision.
• Jobs with more exceptions than rules. If half the steps begin with “but this customer is different”, the work itself needs simplifying first. Otherwise the tool stops at every exception.
• Jobs whose data is scattered or inconsistent. If the same information sits in three files in three different formats, the first stage of automation is bringing the data together in one place.
• Jobs done very rarely. For a job done once a year, a well-written checklist is often enough.

A quick diagnostic table

A common mistake: starting with the hardest job

Companies usually want to start automation with the job they complain about most. It is an understandable wish; but the job people complain about most is often the most complex one: full of decisions, full of exceptions, involving more than one department. When the first tool is built for a job like that, the project drags on, expectations drop and the team concludes that “this sort of thing is not for us”.

A better start is a small job that comfortably meets all three tests. For example, a data entry task that takes up half a day of one person's time every week. A job like this becomes a tool in a few weeks, the hours saved become visible and the team learns to use the tool. By the second job, the data is tidier and the team knows what is possible.

Then there is the question of measurement. Before the tool is set up, note down how long the job takes today: how many hours a week, how many people, how many error corrections a month. Without these figures it is hard to show what has changed once the tool is running.

Before turning a job into a tool

If you have found a candidate job, preparing three things before we talk shortens the process: the job's current steps (who, from which file, to where), a sample input file and an example of the result it should produce. With these three on the table, what the tool will do becomes clear in the first meeting. We look separately at when Excel stops being enough in When does a spreadsheet stop being enough?

In short

A job suited to automation is one that repeats often, whose rules can be written down and whose data is digital. Decisions, exceptions and scattered data leave that part of the job with people, or call for preparation first. You can see how the tools we have built for companies in different industries work on the Our work page.

Frequently asked questions

What is the quickest way to tell whether a job is suited to automation?

Ask the person who does it to write it down step by step, as if explaining it to a new starter. If the steps fit on half a page and the phrase “it depends, I take a look” rarely comes up, the job is very likely a good fit.

Our data is in Excel, but it is messy. Is automation still possible?

Usually yes, but the first job is tidying the data: the same information in the same column, in the same format. The tidying step often brings as much benefit as the automation itself.

Does automation take the work away from people entirely?

Usually not. The repetitive part moves to the tool; checking, approval and exceptions stay with people. A well-built tool frees up a person's time for the part that needs judgement.