The cost of manual work rarely shows up as a line you can point at in the accounts. It shows up in the working day: half an hour on Monday morning preparing payment reminders, twenty minutes each evening copying the day's orders, a month-end morning spent rebuilding an export for the accountant.
Few people decide to work this way. These tasks build up one at a time, each too small to justify a project of its own, until they take up a share of the week that nobody has really costed.
Where they pile up in a small company
- The quote accepted in one tool, then retyped into the invoice or the management system, same lines and same risk of divergence.
- Payment reminders sent when somebody remembers, from a tracking sheet that is updated irregularly.
- Expense receipts photographed, then keyed in a second time for the accounts.
- The monthly export for the accountant, rebuilt every time because the expected format matches none of your tools.
- Orders arriving by email or phone, copied into a spreadsheet before they reach fulfilment.
- The job schedule kept in a shared file that three people edit at once.
Three good candidates to start with
Start with what is both frequent and regular in shape. Turning an accepted quote into an invoice is often a good candidate: the data already exists, it is structured, and a copying error can have a direct cost.
Payment reminders are another good candidate. The point is not to send the same automatic message to everyone, but to know without having to think about it who is late and by how long, then let a person choose the tone and the next step.
The exchange with the accountant is a third candidate: fixed format, fixed date, rebuilt by hand every month. Because the need comes round at regular intervals and the expected format barely moves, the time spent is easy to measure and to compare against the cost of automating it.
First, check the tool you already pay for
Before commissioning anything, open the settings of the software you already pay for. A share of the manual work in a small company maps to a feature nobody switched on, a document template never created, or a form field left optional when it should be required.
That check is unglamorous, it takes very little time, and it sometimes saves you from commissioning something the tool already does.
What is better left to people
- Anything that is a commercial gesture: a discount, an extended deadline, a client you deliberately do not chase this week.
- Rare cases that differ every time. If they take little time in total and none of them resembles the last, handling them by hand can cost less than encoding every exception.
- A process you already know is about to change, because you would pay for the same work twice.
- Decisions whose errors surface late. Build the check that catches them first.
Measure before, then after
Take the task you suspect is worst and count it over a representative period: a mark each time, a rough duration, the number of people involved. For steady work a week already gives you an order of magnitude; for seasonal or month-end work, measure across that cycle. Then convert those hours into money at a realistic loaded rate.
Measure again a few weeks after it goes live, using the same method. That tells you whether the estimate held, and it is the figure that makes the next decision easier to make.