Guide

Working out the ROI of an automation

AgenIA · August 2026 · 5 min read

Automating a task costs something: time to set up, sometimes a subscription, always a little adjustment of habits. Before committing, it is worth knowing whether the candle is worth the game. Here is a simple method, with no complicated spreadsheet, for estimating whether an automation pays for itself — and how quickly.

The question to ask before calculating anything

Not every repetitive task is a good candidate for automation. The question is not just "is this repetitive?" but "is this repetitive and predictable?". A task that always follows the same shape, with few exceptions, lends itself well to the calculation below. A task that comes out differently every time needs clarifying first, otherwise the ROI figure is wrong from the outset.

The basic formula

The annual time saved by an automation comes from three numbers you already have, or can estimate in a few minutes:

1. Time per occurrence. How many minutes the task takes each time it is done by hand.
2. Annual frequency. How often the task comes round in a year (per day × working days, for instance).
3. Fully loaded hourly cost. The real cost of an hour of the time of the person doing it — total employment cost, not take-home pay.
4. Set-up cost. The time or budget needed to automate the task, converted into money.

The calculation then fits on one line:

Annual gain = (time per occurrence × annual frequency × hourly cost) − set-up cost

A worked example

Take chasing unpaid invoices by hand: it takes 8 minutes and comes round 3 times a week (roughly 150 times a year), at a fully loaded hourly cost of €30:

1. Annual time spent. 8 minutes × 150 = 1,200 minutes, or 20 hours a year.
2. Annual cost of the manual task. 20 hours × €30 = €600 a year.
3. Estimated set-up cost. An automatic chasing sequence usually means a few hours of configuration — say €400 all in for the first year.
4. Net gain in year one. €600 − €400 = €200, then €600 a year afterwards, with no further set-up cost.

That figure alone does not say everything — it ignores, for instance, the money recovered faster thanks to more regular chasing — but it gives a simple decision threshold you can check against your own numbers.

The costs people forget to include

• Time spent correcting mistakes. A manual task produces errors that cost time elsewhere (an invoice mis-keyed, throwing off a bank reconciliation, for example). That correction time is part of the real cost of doing it by hand.
• Waiting time. A task that depends on someone available only at certain moments (approval, sending) slows down everything downstream. That delay has a cost, even though it appears on no payslip.
• Maintenance. An automation is not free once it is live — it needs a minimum of attention when the tools or the processes change. Build that in, even modestly.

A simple threshold for deciding

In practice, an automation is worth launching as soon as the net gain in year one is positive and the recurring annual gain (in the years after) clearly exceeds the estimated maintenance cost. If the calculation comes out close to zero, it is not necessarily a bad idea — but it is probably not the priority next to another task with a better ratio.

Three common mistakes

1. Understating the fully loaded hourly cost. Many smaller companies think in terms of net pay, which artificially shrinks the real gain from an automation.
2. Counting only the time, never the errors avoided. The minimal calculation above is already useful, but it often understates the true value of a reliable automation.
3. Automating the most visible task rather than the most profitable one. The task that irritates everyone day to day is not always the one with the best gain-to-cost ratio — the calculation settles it objectively.

Read next

• Audit your processes before you automate — what to have looked at before putting numbers on anything.
• 20 tasks to automate first — the candidates that most often pay back.
• Take a look at your processes — thirty minutes, no obligation.