Which jobs in your business are worth automating?

Start with where the time goes

Most businesses can name the jobs they hate. Fewer can say how long those jobs actually take.

Before deciding anything, spend one normal week writing down every repetitive task and roughly how long it took. Don’t tidy it up. “Copied orders from email into the system, 2 hours” is exactly the level of detail you want.

At the end of the week, multiply by four for a monthly figure. That list is the raw material for everything below.

The five-question test

Run each job on your list through these questions. The more yeses, the better the candidate.

1. Does it happen often?

Daily is good. Several times a day is better. A job you do once a month rarely repays the cost of building something, however tedious it is.

2. Does the information arrive in a similar shape each time?

Orders, bookings, invoices, enquiry forms and timesheets usually do. The wording might vary, but the same details turn up: a name, a date, a product, a quantity. That consistency is what a system needs.

3. Could someone write down the rules?

If you could hand a new starter a one-page sheet explaining how to do the job, a system can probably follow the same rules. If the honest answer is “it depends, you just get a feel for it”, that’s a sign the job needs a person.

4. Do mistakes cost money?

Retyping is where typos come from. A wrong quantity or a transposed postcode means a wrong delivery, a refund or an awkward phone call. Jobs where errors are expensive are worth automating even at lower volumes, because a system makes the same check every time.

5. Is it moving information from one place to another?

A surprising amount of office work is copying: from email into a system, from one spreadsheet into another, from five tools into one report. Copying is what software is best at.

Jobs that usually pass

Common examples:

Order and booking emails that someone retypes into another system. (How that works.)
Monthly reports stitched together from several tools. (One report from many tools.)
Website pages built from information you already hold, like product details or locations. (Doing that without it reading like spam.)
Chasing and reminders: invoices, renewals, missing paperwork.
Sorting incoming email so the right person sees the right message first.

The biggest one I run read 7,369 order emails for a UK manufacturer in the last 30 days: sorting them, reading the orders and queuing them by priority. It passed all five questions easily. The case study has the details.

Jobs that usually fail

Rare jobs. Year-end tasks, one-off projects, anything you do a handful of times.
Broken processes. If the job is hard because nobody agrees how it should be done, automating it just makes the disagreement faster. Fix the process, then decide.
Judgement as the product. Pricing a complicated quote, handling a complaint, deciding who to hire. A system can gather the information and draft something, but a person should decide.
Jobs that are about to change. If you’re switching systems in six months, wait.

Put a number on it

Once you have a shortlist, turn the hours into money:

hours per month × cost per hour × 12 = what the job costs you a year

Compare that with the cost of automating it over the same year. What AI automation costs has the typical prices and a worked example.

A job that takes an hour a week is rarely worth a build on time alone. A job that takes an hour a day might well be, especially if mistakes in it are expensive.

Freeing people up, not replacing them

The point of all this isn’t fewer people. It’s the same people spending less time copying and more time on the work that actually needs them: talking to customers, solving problems, improving things.

When the copying goes, those hours don’t disappear. They go back to your team.

If you’d rather not do this yourself

This is what an audit is for. I look at where the hours go, test each job against questions like these, and give you a written report of what’s worth automating and what it would cost.